+ All Categories
Home > Documents > Research Article Vehicle Safety Enhancement System...

Research Article Vehicle Safety Enhancement System...

Date post: 20-Mar-2018
Category:
Upload: ngothuy
View: 215 times
Download: 0 times
Share this document with a friend
20
Hindawi Publishing Corporation International Journal of Distributed Sensor Networks Volume 2013, Article ID 542891, 19 pages http://dx.doi.org/10.1155/2013/542891 Research Article Vehicle Safety Enhancement System: Sensing and Communication Huihuan Qian, 1,2 Yongquan Chen, 1 Yuandong Sun, 1 Niansheng Liu, 3 Ning Ding, 1 Yangsheng Xu, 1,2 Guoqing Xu, 1,2 Yunjian Tang, 4 and Jingyu Yan 1,2 1 Department of Mechanical and Automation Engineering, e Chinese University of Hong Kong, Shatin, Hong Kong 2 Shenzhen Institute of Advanced Technology, Shenzhen, Guangdong, China 3 School of Computer Engineering, Jimei University, Xiamen, Fujian, China 4 Chongqing Research Center for Information and Automation Technology, Chongqing Academy of Science and Technology, Chongqing, China Correspondence should be addressed to Guoqing Xu; [email protected] Received 11 April 2013; Accepted 4 October 2013 Academic Editor: Liusheng Huang Copyright © 2013 Huihuan Qian et al. is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. With the substantial increase of vehicles on road, driving safety and transportation efficiency have become increasingly concerned focus from drivers, passengers, and governments. Wireless networks constructed by vehicles and infrastructures provide abundant information to share for the sake of both enhanced safety and network efficiency. is paper presents the systematic research to enhance the vehicle safety by wireless communication, in the aspects of information acquisition through vehicle sensing, vehicle- to-vehicle (V2V) routing protocol for the highly dynamic vehicle network, vehicle-to-infrastructure (V2I) routing protocol for a tradeoff in real-time performance and load balance, and hardware implementation of V2V system with on-road test. Simulations and experimental result validate the feasibility of the algorithms and communication system. 1. Introduction Safety-on-road is an important concern by all the drivers and passengers. With the sharp increase of vehicles on road, this has become increasingly important and challenging due to the higher probability of accidents. ere are two aspects which can help prevent the occurrence or the deterioration of an accident. (1) Notify the immediately following vehicles to prevent chained accidents. However, bad weathers, such as fog, will prevent the drivers from obtaining the accident occurrence on time if merely from vision. (2) Notify other vehicles which may take the path encountering the accident spot, so as to avoid traffic jam. However, the current radio- based broadcasting approach requires human interaction for both accident detection and broadcasting and thus cannot achieve satisfying real-time performance. erefore, a real- time vehicle safety enhancement system is needed for col- lision detection and vehicle status information sharing with nearby vehicles. Two subsystems are necessary in the vehicle safety enhancement system. One subsystem, vehicular information acquisition subsystem (VIAS), takes charge of vehicular status sensing, so as to detect the accident information (e.g., position, etc.), classify the severity level (e.g., slight bumping, rolling over, etc.), as well as to detect some hazardous driving behaviors (e.g., sudden slow down, S-shape driving, etc.). Current additive sensor technologies, (e.g., GPS, accelerom- eter, angular sensors, etc.), as well as the onboard vehicular signals (e.g., airbag deploying signal, brake pedal signal, vehicle speed signal, etc.), make possible the implementation of this subsystem. e other subsystem, intervehicle wireless communica- tion subsystem (IVWCS), is in charge of the intervehicular information sharing. Recent advances in wireless technolo- gies have made vehicle-to-vehicle (V2V) communications and vehicle-to-infrastructure (V2I) communication possible by means of mobile ad hoc networks (MANETs) for intelli- gent transportation system (ITS). Vehicle ad hoc networks
Transcript
Page 1: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2013 Article ID 542891 19 pageshttpdxdoiorg1011552013542891

Research ArticleVehicle Safety Enhancement SystemSensing and Communication

Huihuan Qian12 Yongquan Chen1 Yuandong Sun1 Niansheng Liu3 Ning Ding1

Yangsheng Xu12 Guoqing Xu12 Yunjian Tang4 and Jingyu Yan12

1 Department of Mechanical and Automation Engineering The Chinese University of Hong Kong Shatin Hong Kong2 Shenzhen Institute of Advanced Technology Shenzhen Guangdong China3 School of Computer Engineering Jimei University Xiamen Fujian China4Chongqing Research Center for Information and Automation Technology Chongqing Academy of Science and TechnologyChongqing China

Correspondence should be addressed to Guoqing Xu gqxumaecuhkeduhk

Received 11 April 2013 Accepted 4 October 2013

Academic Editor Liusheng Huang

Copyright copy 2013 Huihuan Qian et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

With the substantial increase of vehicles on road driving safety and transportation efficiency have become increasingly concernedfocus from drivers passengers and governments Wireless networks constructed by vehicles and infrastructures provide abundantinformation to share for the sake of both enhanced safety and network efficiency This paper presents the systematic research toenhance the vehicle safety by wireless communication in the aspects of information acquisition through vehicle sensing vehicle-to-vehicle (V2V) routing protocol for the highly dynamic vehicle network vehicle-to-infrastructure (V2I) routing protocol for atradeoff in real-time performance and load balance and hardware implementation of V2V system with on-road test Simulationsand experimental result validate the feasibility of the algorithms and communication system

1 Introduction

Safety-on-road is an important concern by all the driversand passengers With the sharp increase of vehicles on roadthis has become increasingly important and challenging dueto the higher probability of accidents There are two aspectswhich can help prevent the occurrence or the deteriorationof an accident (1) Notify the immediately following vehiclesto prevent chained accidents However bad weathers suchas fog will prevent the drivers from obtaining the accidentoccurrence on time if merely from vision (2) Notify othervehicles which may take the path encountering the accidentspot so as to avoid traffic jam However the current radio-based broadcasting approach requires human interaction forboth accident detection and broadcasting and thus cannotachieve satisfying real-time performance Therefore a real-time vehicle safety enhancement system is needed for col-lision detection and vehicle status information sharing withnearby vehicles

Two subsystems are necessary in the vehicle safetyenhancement system One subsystem vehicular informationacquisition subsystem (VIAS) takes charge of vehicularstatus sensing so as to detect the accident information (egposition etc) classify the severity level (eg slight bumpingrolling over etc) as well as to detect some hazardous drivingbehaviors (eg sudden slow down S-shape driving etc)Current additive sensor technologies (eg GPS accelerom-eter angular sensors etc) as well as the onboard vehicularsignals (eg airbag deploying signal brake pedal signalvehicle speed signal etc) make possible the implementationof this subsystem

The other subsystem intervehicle wireless communica-tion subsystem (IVWCS) is in charge of the intervehicularinformation sharing Recent advances in wireless technolo-gies have made vehicle-to-vehicle (V2V) communicationsand vehicle-to-infrastructure (V2I) communication possibleby means of mobile ad hoc networks (MANETs) for intelli-gent transportation system (ITS) Vehicle ad hoc networks

2 International Journal of Distributed Sensor Networks

(VANETs) considering the special features of vehicularenvironment have clear benefits that is VANETs not onlyimprove the overall safety for vehicular system but alsosmooth the traffic flow

Furthermore in order to disseminate the vehicle statusinformation to the traffic center communication channelsshould be established The current approaches utilize mobilephone channels such as GSM GPRS or 3G and so forthHowever the numerous vehicles on road will lay a tremen-dous burden on the mobile phone network Hence therehave been proposals to establish a series of infrastructuresalong the road as information sinking ports and finally formthe V2I network [1] Nevertheless before this giant scope ofconstructions the robotic concept of mobile nodes can serveas sinking ports for establishing the overall car safety networkframework as a transitional and testing step

Governments and prominent industrial corporations areinvolved in this field over the past ten years such as GMBMW and Toyota Several important projects were held topromote development of such VANETs system AdvancedDriver Assistance Systems (ADASE2) [2] and Crash Avoid-ance Metrics Partnership (CAMP) [3] were launched inthe US The project ldquoFleetNetmdashInternet on the roadrdquo wasset up in German [4] ldquoFleetNetrdquo uses FleetNet networklayer (FNL) to realize a vehicular ad hoc network It isan adaptive protocol and position information are usedfor routing and forward strategy Taleb et al proposed ascheme called ROMSGP to enhance the stability of the IVCand RVC communications in VANETs the key idea is togroup vehicles according to their moving directions [5]Moreover the Federal Communications Commission hasallocated spectrum and new standards for VANETs morerecently called IEEE 80211p IEEE 80211p is an approvedamendment to the IEEE 80211 standard to addwireless accessin vehicular environments (WAVE) It defines enhancementsto 80211 required to support intelligent ITS applications

Although the need for V2V communication is evidentand many studies for VANETs have been done so far fewintervehicle communication systems for information sharingbetween vehicles have been put into operationOnStar systemis designed for vehicle safety byGeneralMotors [6] It is basedon wireless cellular communications such as GSM GPRS or3G This system works well but it limits the simple servicebetween the driver and the OnStar Center when the crashaccident happens In addition the service is available only forall vehicles that have the factory-installed OnStar hardwarelike GM vehicles

This paper proposes a vehicle safety enhancement systembased on wireless communication It takes into considerationthe vehicular information acquisition communication issuesfor V2V and V2I and finally the hardware implementationThis paper is organized as below Section 2 introduces thevehicular information acquisition subsystem which collectssignals for vehicle accident detection and classification so asto share with the whole vehicle networks In Section 3 weselected the benchmark routing protocol for V2V networkby comparison of three classical protocols that is DSDVDSR and AODV In Section 4 we consider the uniquefeatures of vehicle networks that is with availability of

position and velocity and modify the adopted AODV intoGPS based AODV (GBAODV) which has been verified forbetter performance through simulation Section 5 addressesthe problem of vehicle-to-infrastructure (V2I) informationsinking in which we consider the optimization of time-delayand network load In Section 6 we describe the real-timeexperiment Section 7 concludes the paper

2 Vehicular Information Acquisition [7]

Traffic collisions are serious threats to human lives It isestimated that 12 million people were killed in motor-vehicleaccidents globally [8] There are various factors such asroad and traffic environment vehicle states and driversrsquobehaviors Vehicular information if obtained and sharedamong surrounding vehicles can reduce the risk of accidents

21 Collision and Potentially Hazardous State ClassificationWe classify and define the accidents or potential hazardousactions into 4 severity levels as listed below

(1) Level 1 (potentially hazardous) sudden slow down[9]

(2) Level 2 (minor accident) slight bumping [10]

(3) Level 3 (major accident) airbag deployed [11]

(4) Level 4 (serious accident) rolling over [12]

Different severity levels should be processed differently Ifsudden slowdownoccurs the immediately following vehiclesought to be notified so as to prevent collision For a minoraccident a flexible traffic police for example motorcyclepolice will be adequate and the most urgent need is torecover the traffic If the airbag is activated or there isrolling over ambulance should be connected automaticallyand promptly to save the injured Even more for the rollingover case a crane may also be needed to remove the vehiclein accident

22 Sensors for Collision Detection During the driving somereal-time status signals are available inside the vehicle such asairbag deployment signal Aside of those additional sensorsare required to classify the signals into more delicate levelsThese signals and sensors are listed in Table 1

23 Sensor Data Fusion and Collision Estimation We adoptthe Freescale9s12XEP100 as the coreMCU controller for datagathering and analysis It is configured with 16 bit HCS12XCPU 512 Kbyte Flash EEPROM and 32Kbyte RAM fordigital signal processing

According to the sensor information obtained fromvehicle the real-time estimation of collision is conducted asshown in Figure 1

Figure 2 illustrates the overall processing action flowchartfor collision detection and dissemination

International Journal of Distributed Sensor Networks 3

Table 1 Sensors for vehicular information acquistion

Sensor FunctionAirbag deployment LED(on dashboard) Airbag deployment detection

Vehicle wheel sensor Detect vehicle speedGPSSiRF starIII Vehicle localization6G one-axis accelerometer Sudden slow down detection500G one-axisaccelerometer Slight bumping detection

15 G three-axisaccelerometer Rolling over

Steering wheel angularsensor S-shape steering detection

Slight bumping2

15G accelerometer

reading abnormal500G accelerometer

reading abnormal6G accelerometer

reading abnormal

Airbag deployed

Sudden slowdown1

Airbag deployed3

Level =

Level =

Level =

Level =

Airbag deployed4

Level = 0

Yes

Yes

Yes

Yes

No

No

No

No

Figure 1 Flowchart for collision estimation

3 Evaluation of Classical RoutingProtocols in MANET [13]

The speed of vehicle moving on a freeway is fairly fast atapproximately 80 kmh in general When the accident occurson a freeway the warning message must be instantly sentto other vehicles nearby via the VANET so that the driversnearby have enough time to avoid collision In order toachieve reliable real-time communication the performanceof routing protocol used by the VANET is important [5 14]Different routing protocols have different network charac-teristics [15ndash17] VANET on the freeway is a fast and highlydynamicmobile ad hocwireless networkwithout fixed routerhosts or wireless based stations A significant challenge in thefreeway VANET is how to select an efficient routing protocolamong numerous protocols Thus three classical routing

Real-time vehicle statemonitoring

Real-time estimation ofcollision

Is vehiclecondition safe

Severity level = 1

Report to traffic centerand forward to nearby

vehicles

Forward waning tonearby vehicles

through VANET

Yes

Yes

No

No

Figure 2 Algorithm flowchart of the collision evaluation

protocols are evaluated and compared A proper routingprotocol will be selected based on which modification canbe made to bear the unique features for VANET

31 Classical Routing Protocols for MANET Numerous adhoc routing protocols have been proposed by the InternetEngineering Task Force (IETF) Mobile Ad hoc Networks(MANET) Working Group [18] These routing protocols aredivided into two major classes based on the underlyingrouting information update mechanism employed reactive(on-demand) or proactive (table-driven) For performanceevaluation we have compared three typical routing protocolsin MANET for V2V application Destination sequenceddistance vector protocol (DSDV) is selected as an example fortable-driven protocols while both dynamic source routing(DSR) and ad hoc on-demand distance vector (AODV)protocols are selected as examples for on-demand protocols[19ndash21]

(1) DSDV DSDV is based on the Bellman-Ford algorithmwhich can effectively solve routing loop problem [19] Eachnode has a routing table which contains the shortest path toevery other node in the network Each entry in the routingtable contains a sequence number The number is generatedby the destination If a node receives new informationit consults the routing table and uses the latest sequencenumber to forward If the sequence number is the same asthe one existing in the table the route with the bettermetric isused Stale entries are deleted by regular update of its routingtables DSDV is suitable for creating a small-scale ad hocnetwork

(2) DSR DSR uses source routing instead of hop-by-hoppacket routing in comparison with DSDV and has two majorphases which are route discovery and route maintenance[20] Route discovery is used to set up a route from sourcenode to destination by sending RouteRequest packet in

4 International Journal of Distributed Sensor Networks

Figure 3 Simulation scenario of vehicles moving on a freeway

the source node If a node in the path moves away and breakswireless communication route maintenance will rebuilda route from source node to destination one by sendingRouteError packet to the node adjacent to broken link Eachdata packet carries corresponding routing informationThusit eliminates the need to periodically flood the network withtable update messages which are required in a table-drivenapproach However DSR does not locally repair a brokenlink The connection setup delay is higher than that of table-driven protocols

(3) AODV AODV is very similar to DSR [21] It sets up aroute to the destination by sending a RouteRequest messageThe source node and the intermediate nodes store the nexthop information corresponding to each flow for data packettransmission The major difference between AODV andother reactive routing protocols is that it uses a destinationsequence number (DesSeqNum) to find the latest route tothe destination A node updates its path destination only ifthe DesSeqNum of the current packet received is greater thanthe last DesSeqNum stored at the node However AODVrequires more time to set up a connection than some otherapproaches [15 17 21]

32 Simulation Environment and Performance EvaluationThe simulator for evaluating three routing protocols is theNetwork Simulator (NS2 version 233) NS2 provides sub-stantial support for simulation of wireless networks usingdiscrete-event mode

(1) Basic Scenario The simulation scenario is designedaccording to the normal state of car running on a freewayshown in Figure 3 Assume that the freeway is 30meters wideand 500 meters long 40 vehicles are randomly distributedto the four bidirectional lanes of the freeway Each vehicleis regarded as a mobile node moving forward in a randomfashion The maximum velocity of nodes is 35ms withsimulation period of 200 seconds The channel capacity is

Table 2 Packet loss rate of three routing protocols

Time (second) DSDV () DSR () AODV ()100 1272 068 107200 762 048 048

2Mbs The MAC type is IEEE 80211 The CBR traffic modelis used with data packet size 512 bytes and sending rate160Kbps(2) Performance Evaluation The performance evaluationof routing protocols is based on the measurement of thefollowing parameters [16]Packet Loss Rate () Packet loss occurs when one or morepackets of data traveling across a VANET fail to reach theirdestinations The packet loss rate is calculated as

119901 = (1 minus119873119903

119873119904

) times 100 (1)

where 119873119903and 119873

119904represent the number of data packet

received and sent respectivelyEnd-to-End Delay (ms) End-to-end delay is defined as thetime taken for a data packet to be transmitted across thenetwork from the source to the destinationPacket Jitter (ms) Packet jitter is defined as the delay variationbetween two consecutively received packets belonging to thesame stream In general it is expressed as an absolute valueof delay variationThroughput (Kilobits per Second) System throughput is thesum of the data rates that are delivered to all nodes in thenetworkProtocol Overhead Protocol overhead refers to the number ofroutingmessages requestedwhen a data packet is successfullydelivered to the destination

33 Simulation Results and Discussion Throughout the sim-ulation we discuss the results as below

(1) Packet Loss Rate Table 2 shows the packet loss rateincurred by DSDV AODV and DSR Both DSR and AODVon-demand routing schemes have considerably less packetloss rate than DSDV Because each node is fast mobile in arandom fashion the network topology often changes Thesepackets sent may be lost once the routing table is not timelyupdated in DSDV This results in a higher packet loss rate ofDSDV

(2) End-to-End Delay Figure 4 shows the end-to-end delayof data packets The statistical values of them are (7440 plusmn

4157)ms (DSDV confidence level 120572 = 095) (8167 plusmn

9210)ms (DSR 120572 = 095) and (8107 plusmn 5026)ms (AODV120572 = 095) respectively They have no significant differencein the mean values However as a whole the margin of delayfluctuation in DSR is the highest among the three protocolsIn DSR a neighbor displacement is noticed only after apacket is sent explicitly to that node The network reacts

International Journal of Distributed Sensor Networks 5

0

001

002

003

004

005

006

007

Pack

et d

elay

(s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

0 1000 2000 3000 4000 5000 6000 70000

005

015

01

025

02

035

03

Packet ID

Pack

et d

elay

(s)

(b) DSR

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

008

Packet ID

Pack

et d

elay

(s)

(c) AODV

Figure 4 Packet delay of three routing protocols

if an acknowledgement is not received Consequently thisincreases packet delay since the packet must wait until a newroute is established(3) Packet Jitter Figure 5 shows the jitter of data packetsusing three different routing protocols The statistical valuesof them are (2356 plusmn 4199)ms (DSDV 120572 = 095) (3254 plusmn

10630)ms (DSR 120572 = 095) and (4498 plusmn 4498)ms (AODV120572 = 095) respectively DSDV shows the most superioritythan others The average jitter of AODV is the biggest amongthe three protocols Moreover the standard deviation of jitterin AODV is larger than that in DSDV Although the averagejitter of DSR is not the largest among the three protocolsDSR has the largest jitter fluctuation among themThismeansviolent variation exists in the delay of a few data packets sentin DSRThe rapid change of jitter is attributed to the frequentchange of network topology and the mechanism of inherentrouting update(4) Throughput Figure 6 shows the throughput compari-son of DSDV DSR and AODV All throughputs are ever-increasing with the time in general This can be attributed

to more active nodes to join network communication withtime extensionThe graph also reveals that AODV has higherthroughput than DSR In DSR a route is chosen basedon the short delay at the instance of route establishmentAlthough this path may be the best route at that instantit may be also a route that lacks routing stability or hasunacceptably high load In contrast AODVhas amore robustupdate mechanism to avoid bottleneck and congestion andeventually improve throughput In DSDV high packet lossrate causes throughput to drop

(5) Protocol Overhead Table 3 shows the overhead compar-ison of DSDV AODV and DSR AODV has the highestoverhead among three routing protocols due to three mainreasons Firstly AODV allows broadcast Although the dis-covery packets are broadcasted only when necessary suchas establishing a new route link breakage or route errorthe broadcast instances will often appear for a fast mobileVANET Secondly AODV allows mobile nodes to respondto link breakages and changes in network topology in a

6 International Journal of Distributed Sensor Networks

0

001

002

003

004

005

006

007

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

0

005

015

01

025

02

035

03

(b) DSR

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

Packet ID

(c) AODV

Figure 5 Packet jitter of three routing protocols

timely mannerThus numerous routingmessages are used tomaintain an active route in AODVThirdly AODV is derivedfrom DSDV It still has similar features of proactive routingprotocols When the network topology is often changedbecause of the fast mobility of nodes proactive protocolsmust send more messages to maintain a valid routing tableThe experimental results are shown in Table 3 DSR hasthe best protocol overhead performance among the threeprotocols The DSR protocol is composed of the two mainmechanisms of ldquoroute discoveryrdquo and ldquoroute maintenancerdquowhich work together to allow nodes to discover andmaintainroutes to arbitrary destinations in the ad hoc network Somemeasures reducing overhead are adopted in the process ofldquoRoute Discoveryrdquo and ldquoRoute Maintenancerdquo of DSR [21]

34 Conclusion Based on the simulation results we discov-ered that DSDV achieves marginal better packet delay andpacket jitter than AODV and DSR However it has signif-icantly higher packet loss rate and the smallest throughoutamong them Although DSR has the best protocol overhead

Table 3 Overhead of three routing protocols

Type DSDV AODV DSRNumber of received data packets 5579 5999 5993Number of sent message packets 25607 39488 1615Protocol overhead 459 658 027

performance among three routing protocols it has poorerjitter and throughput thanAODV As a whole AODV ismoreappropriate for the freeway VANET according to the qualityof service and the real time of packet delivery The inherentreason producing this result is the effect of node mobility onthe performance of the routing protocol

4 GPS Based AODV Protocol forV2V Communication [22]

As we can find through the previous protocol evaluationAODV is appropriate for the dynamic structure with mobilevehicles The source node will find a route to destination

International Journal of Distributed Sensor Networks 7

0 50 100 150 2000

20

40

60

80

100

120

140

Time (s)

Thro

ughp

ut (k

bps)

DSRDSDVAODV

Figure 6 Throughput comparison of three routing protocols

for data transmission To find a route to the destination thesource broadcasts a route request packet (RREQ) The nodereceiving RREQ will reply a route reply packet (RREP) to thereverse path that RREQ went through if it is the destinationof RREQ or if it has a recent route to the destinationOtherwise the node will broadcast RREQ until either ofthe two situations above occurs As RREP traverses backto the source the nodes along the path enter the forwardroute into their routing tables If a node leaves the networkthe node that discovers this broken link will send a linkfailure notification (RERR) to the precursorsThe RERR goesupstreamuntil it reaches the sourceThe sourcewill thereafterrestart route discovery process if needed [23]

However AODV is dedicated forMANET andmodifica-tion is necessary for VANET applications [24] Since nodes inVANET are fast moving vehicles the topology of the networkchanges all the time making the position and velocity ofthe node critical factors while finding the routes Thereforebesides conventional topology-based routing protocol (suchas AODV) position-based routing protocol which primarilyuses the position information obtained byGPS to find a routeis applied in VANET [25]

In recent years many researchers have tried to combinetwo types of routing together In [26] Kim et al pro-posed AODV-RRS which restricts the number of forward-ing RREQs according to the concepts of stable zone andcaution zone PAODV [27] restricts the number of floodingRREQs based on the distance between current node andits neighbors Both of them have similar mechanism andreduce the number of broken links DAODV [28] establishesa route depending on the direction and position of sourceintermediate and destination node Although it also reducesthe number of broken links DAODV assumes that sourcenode knows the direction and position of all the nodes inthe network Actually this is impossible with the current GPS

devices We should find some mechanisms to get the geo-graphic information of all the nodesThereby in [29] Asenovand Hnatyshin proposed GeoAODV routing protocol Eachnode will maintain an additional table geotable to keep trackof the geographic information of all the nodes RREQs arerestricted flooding in the region determined by the geotableHowever it costs more resources to maintain two tables

We go a step further in this section by proposing aGPS based AODV (GBAODV) which enhances the overallperformance of AODV in VANET In order to be attainablein physical implementation GBAODV assumes that eachnode only knows its own position and velocity Withoutmaintaining a geotable [29] GBAODV is muchmore concisethan GeoAODV

41 Overview of GBAODV With current GPS device wecan obtain the longitude (119909 coordinate) and latitude (119910coordinate) of current node and calculate the speed in eachdirection with two successive sets of coordinates Withoutmaintaining a geotable we add geographic information intorouting table to make the algorithm concise

There are two main features in GBAODV First is toreduce the number of RREQs The node receiving an RREQpacket will check the distance and motion trend betweenthe precursor and itself so as to decide if this RREQ shouldbe broadcasted By restricting these flooding RREQs it willavoid lots of packet collisions in the network As a resultthe packet delivery ratio and throughput of the networkcan be raised Second is to mark the route according to thepositions and velocities of source intermediate and desti-nation Higher mark means higher stability Consequentlyevery node will choose a route with higher mark

These two features require specifying flooding rules andmarking standards as well as some modifications in therouting table and routing packets

42 Modified Structure of Routing Table RREQ and RREP

(1) Modified Routing Table Original routing table containsthe IP addresses of next hop and destination sequencenumber of destination and total hops The routing table willupdate if the sequence number of destination is larger ortotal hops is less We added the position and velocity of thedestination node and the mark of this route into the routingtable Besides the two conditions above the routing table willupdate when the mark is larger which means that the routepath is more stable

(2) Modified Frame Structure of RREQ We inserted position(119909 119910 coordinates) and velocity (119909 119910 direction) of the sourceand current node into RREQ packet We also applied areserved segment to record the mark of the route link fromsource to precursor

(3) Modified Frame Structure of RREP RREP is modifiedsimilar to RREQ We inserted position and velocity of thedestination and current node into RREP packet We also

8 International Journal of Distributed Sensor Networks

P

B

A

C

DE

04 lowast R

08 lowast R

Figure 7 B C and D will broadcast RREQ received from P

applied a reserved segment to record the mark of the routefrom destination to current node

43 Flooding Rules and Marking Standards

(1) Flooding Rules Let 119889 stand for the distance between theprecursor and current node and 119877 stand for the maximumtransmission distance of the wireless cardThe flooding rulesare as follows

If (119889 lt 04 119877 (119889 gt 08119877ampamp119889 is increasing))

discard RREQ

Else

broadcast RREQ

The coefficients 04 and 08 are chosen empiricallyWe conducted simulations with the combinations of thesecoefficients to study numbers of RREQs and RERRs packetloss ratio and average end-to-end delay Results show that thecombination of 04 and 08 is the best among all

In Figure 7 B C andDwill broadcast the RREQ receivedfrom P while A and E will discard it

(2) Marking Standards As illustrated in Figure 8 let 119871119863119888

stand for the current distance between node 119871 and node 119868119871119863119897stand for this distance a moment later 119877119863

119888stand for the

current distance between node 119868 and node 119877 and 119877119863119897stand

for this distance a moment later We specify the markingstandards as follows

mark = 2 sgn (119871119863119888minus 119871119863119897) + 2 sgn (119877119863

119888minus 119877119863119897) + 119862 (2)

where sgn() is a signum function119862 is a constant

119862 = 0 if 119868 is in the dashed rectangleminus3 if 119868 is out of this rectangle

(3)

minus3 is chosen experimentally according to simulationsSince the transmission distance of wireless LAN card is

limited geographic distance of each hop is the key factor thataffects the quality of communication If the distance stays

L

R

I1

I2

Figure 8 Different mark for different regions

unchanged then the route is really stable so that the sourceand destination can communicate steadily If the distance isincreasing it is possible that one node in this routewill exceedthe transmission distance soon If the distance is decreasingthen after some time the two nodes may cross over and startto depart It is better than increasing distance but not asgood as unchanged distance So we set two values for eachchanged distance 0 for each decreasing distance and minus2 foreach increasing distance as shown in (2)

To find a path to the destination the source broadcastsan RREQ containing the position and velocity of source andpreset a standard mark value

As illustrated in Figure 9 when a node receives an RREQthe following will occur

(i) If it does not have a route to the source node it willinsert the route RREQ into its routing table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination or has a route to destination itwill create an RREP in which the mark is a standardvalue (if it is the destination) or is obtained fromrouting table (if it has a route to destination)

(iv) If it is a node besides (3) and satisfies the broadcastingconditions (refer to flooding rules) it will add themark of sourcerarr precursorrarr current to the presentmark in RREQ It will also update the position andvelocity of current node in RREQ

As illustrated in Figure 10 when a node receives an RREPthe following will occur

(i) If it does not have a route to the destination it willinsert the route which RREP goes through into itsrouting table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination of RREP (ie the source ofRREQ) it will insert the route into routing table

(iv) If it is a forwarding node before unicasting RREQit will add the mark of sourcerarr precursorrarr current

International Journal of Distributed Sensor Networks 9

Receives anRREQ

Has a route tosource

Inserts inrouting table

NeedsupdateUpdates

Destination or has aroute to destination

Replies anRREP

Needsforward

DiscardsRREQ

Updates markand current

nodersquos positionand velocity in

RREQ

BroadcastsRREQ

No

Yes

Yes

No

YesNo

No

Yes

Figure 9 Flowchart of processing RREQ

to the present mark in RREP It will also update theposition and velocity of current node in RREQ

44 Simulation Setup

(1) Simulation Tools We chose VanetMobiSim 11 [30] fortraffic simulation This software can generate a traffic flow inthe format suitable forNS2 [31] which can loadGBAODV fornetwork simulation

With VanetMobiSim we import maps from the USCensus Bureau TIGERLine database [32] which includescomplete coverage of the United States Puerto Rico and soforthMoreoverVanetMobiSim supports formultilane roadsdifferentiated speed constraints and traffic light signals atintersections All the vehicles can be set to Intelligent DriverModel with Lane Changing (IDM LC) [33 34] For thesereasons the scenario in traffic layer is quite authentic whichmakes the simulation in network layer reliable

(2) Parameter Settings In our simulation we observed twotypes of traffic models downtown and highway TGR11001[32] (district of Columbia WA) is chosen as downtown

Receives anRREP

Has a route todestination

Inserts inrouting table

NeedsupdateUpdates

Inserts inrouting table

Sourcenode

Update mark andcurrent nodersquosposition and

velocity in RREP

Unicast RREP

NoYes

Yes

No

Yes

No

Figure 10 Flowchart of processing RREP

Table 4 Parameter settings of traffic

Traffic layer Downtown HighwayArea (m2) 1000 times 1000 1000 times 1000

Number of lanes 3 3Maximum number of traffic lights 10 NoneSpeed (kmh) 20sim80 60sim120Simulation time (s) 250 1000

Table 5 Parameter settings of network

Network layer Downtown HighwayMaximum transmission distance (m) 250 250Number of sources 35 17Number of connections 56 26CBR packet size (bytes) 256 512Transmission rate (pkts) 1 2Simulation time (s) 250 1000

map and TGR36001 [32] (Albany county NY) is chosen ashighway map Multilanes and traffic lights are involved Allthe vehicles follow the IDM LC driving model Tables 4 and5 are the parameter settings of traffic and network simulationTraffic flow and CBR (constant bitrate) data flow are bothgenerated randomly

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 2: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

2 International Journal of Distributed Sensor Networks

(VANETs) considering the special features of vehicularenvironment have clear benefits that is VANETs not onlyimprove the overall safety for vehicular system but alsosmooth the traffic flow

Furthermore in order to disseminate the vehicle statusinformation to the traffic center communication channelsshould be established The current approaches utilize mobilephone channels such as GSM GPRS or 3G and so forthHowever the numerous vehicles on road will lay a tremen-dous burden on the mobile phone network Hence therehave been proposals to establish a series of infrastructuresalong the road as information sinking ports and finally formthe V2I network [1] Nevertheless before this giant scope ofconstructions the robotic concept of mobile nodes can serveas sinking ports for establishing the overall car safety networkframework as a transitional and testing step

Governments and prominent industrial corporations areinvolved in this field over the past ten years such as GMBMW and Toyota Several important projects were held topromote development of such VANETs system AdvancedDriver Assistance Systems (ADASE2) [2] and Crash Avoid-ance Metrics Partnership (CAMP) [3] were launched inthe US The project ldquoFleetNetmdashInternet on the roadrdquo wasset up in German [4] ldquoFleetNetrdquo uses FleetNet networklayer (FNL) to realize a vehicular ad hoc network It isan adaptive protocol and position information are usedfor routing and forward strategy Taleb et al proposed ascheme called ROMSGP to enhance the stability of the IVCand RVC communications in VANETs the key idea is togroup vehicles according to their moving directions [5]Moreover the Federal Communications Commission hasallocated spectrum and new standards for VANETs morerecently called IEEE 80211p IEEE 80211p is an approvedamendment to the IEEE 80211 standard to addwireless accessin vehicular environments (WAVE) It defines enhancementsto 80211 required to support intelligent ITS applications

Although the need for V2V communication is evidentand many studies for VANETs have been done so far fewintervehicle communication systems for information sharingbetween vehicles have been put into operationOnStar systemis designed for vehicle safety byGeneralMotors [6] It is basedon wireless cellular communications such as GSM GPRS or3G This system works well but it limits the simple servicebetween the driver and the OnStar Center when the crashaccident happens In addition the service is available only forall vehicles that have the factory-installed OnStar hardwarelike GM vehicles

This paper proposes a vehicle safety enhancement systembased on wireless communication It takes into considerationthe vehicular information acquisition communication issuesfor V2V and V2I and finally the hardware implementationThis paper is organized as below Section 2 introduces thevehicular information acquisition subsystem which collectssignals for vehicle accident detection and classification so asto share with the whole vehicle networks In Section 3 weselected the benchmark routing protocol for V2V networkby comparison of three classical protocols that is DSDVDSR and AODV In Section 4 we consider the uniquefeatures of vehicle networks that is with availability of

position and velocity and modify the adopted AODV intoGPS based AODV (GBAODV) which has been verified forbetter performance through simulation Section 5 addressesthe problem of vehicle-to-infrastructure (V2I) informationsinking in which we consider the optimization of time-delayand network load In Section 6 we describe the real-timeexperiment Section 7 concludes the paper

2 Vehicular Information Acquisition [7]

Traffic collisions are serious threats to human lives It isestimated that 12 million people were killed in motor-vehicleaccidents globally [8] There are various factors such asroad and traffic environment vehicle states and driversrsquobehaviors Vehicular information if obtained and sharedamong surrounding vehicles can reduce the risk of accidents

21 Collision and Potentially Hazardous State ClassificationWe classify and define the accidents or potential hazardousactions into 4 severity levels as listed below

(1) Level 1 (potentially hazardous) sudden slow down[9]

(2) Level 2 (minor accident) slight bumping [10]

(3) Level 3 (major accident) airbag deployed [11]

(4) Level 4 (serious accident) rolling over [12]

Different severity levels should be processed differently Ifsudden slowdownoccurs the immediately following vehiclesought to be notified so as to prevent collision For a minoraccident a flexible traffic police for example motorcyclepolice will be adequate and the most urgent need is torecover the traffic If the airbag is activated or there isrolling over ambulance should be connected automaticallyand promptly to save the injured Even more for the rollingover case a crane may also be needed to remove the vehiclein accident

22 Sensors for Collision Detection During the driving somereal-time status signals are available inside the vehicle such asairbag deployment signal Aside of those additional sensorsare required to classify the signals into more delicate levelsThese signals and sensors are listed in Table 1

23 Sensor Data Fusion and Collision Estimation We adoptthe Freescale9s12XEP100 as the coreMCU controller for datagathering and analysis It is configured with 16 bit HCS12XCPU 512 Kbyte Flash EEPROM and 32Kbyte RAM fordigital signal processing

According to the sensor information obtained fromvehicle the real-time estimation of collision is conducted asshown in Figure 1

Figure 2 illustrates the overall processing action flowchartfor collision detection and dissemination

International Journal of Distributed Sensor Networks 3

Table 1 Sensors for vehicular information acquistion

Sensor FunctionAirbag deployment LED(on dashboard) Airbag deployment detection

Vehicle wheel sensor Detect vehicle speedGPSSiRF starIII Vehicle localization6G one-axis accelerometer Sudden slow down detection500G one-axisaccelerometer Slight bumping detection

15 G three-axisaccelerometer Rolling over

Steering wheel angularsensor S-shape steering detection

Slight bumping2

15G accelerometer

reading abnormal500G accelerometer

reading abnormal6G accelerometer

reading abnormal

Airbag deployed

Sudden slowdown1

Airbag deployed3

Level =

Level =

Level =

Level =

Airbag deployed4

Level = 0

Yes

Yes

Yes

Yes

No

No

No

No

Figure 1 Flowchart for collision estimation

3 Evaluation of Classical RoutingProtocols in MANET [13]

The speed of vehicle moving on a freeway is fairly fast atapproximately 80 kmh in general When the accident occurson a freeway the warning message must be instantly sentto other vehicles nearby via the VANET so that the driversnearby have enough time to avoid collision In order toachieve reliable real-time communication the performanceof routing protocol used by the VANET is important [5 14]Different routing protocols have different network charac-teristics [15ndash17] VANET on the freeway is a fast and highlydynamicmobile ad hocwireless networkwithout fixed routerhosts or wireless based stations A significant challenge in thefreeway VANET is how to select an efficient routing protocolamong numerous protocols Thus three classical routing

Real-time vehicle statemonitoring

Real-time estimation ofcollision

Is vehiclecondition safe

Severity level = 1

Report to traffic centerand forward to nearby

vehicles

Forward waning tonearby vehicles

through VANET

Yes

Yes

No

No

Figure 2 Algorithm flowchart of the collision evaluation

protocols are evaluated and compared A proper routingprotocol will be selected based on which modification canbe made to bear the unique features for VANET

31 Classical Routing Protocols for MANET Numerous adhoc routing protocols have been proposed by the InternetEngineering Task Force (IETF) Mobile Ad hoc Networks(MANET) Working Group [18] These routing protocols aredivided into two major classes based on the underlyingrouting information update mechanism employed reactive(on-demand) or proactive (table-driven) For performanceevaluation we have compared three typical routing protocolsin MANET for V2V application Destination sequenceddistance vector protocol (DSDV) is selected as an example fortable-driven protocols while both dynamic source routing(DSR) and ad hoc on-demand distance vector (AODV)protocols are selected as examples for on-demand protocols[19ndash21]

(1) DSDV DSDV is based on the Bellman-Ford algorithmwhich can effectively solve routing loop problem [19] Eachnode has a routing table which contains the shortest path toevery other node in the network Each entry in the routingtable contains a sequence number The number is generatedby the destination If a node receives new informationit consults the routing table and uses the latest sequencenumber to forward If the sequence number is the same asthe one existing in the table the route with the bettermetric isused Stale entries are deleted by regular update of its routingtables DSDV is suitable for creating a small-scale ad hocnetwork

(2) DSR DSR uses source routing instead of hop-by-hoppacket routing in comparison with DSDV and has two majorphases which are route discovery and route maintenance[20] Route discovery is used to set up a route from sourcenode to destination by sending RouteRequest packet in

4 International Journal of Distributed Sensor Networks

Figure 3 Simulation scenario of vehicles moving on a freeway

the source node If a node in the path moves away and breakswireless communication route maintenance will rebuilda route from source node to destination one by sendingRouteError packet to the node adjacent to broken link Eachdata packet carries corresponding routing informationThusit eliminates the need to periodically flood the network withtable update messages which are required in a table-drivenapproach However DSR does not locally repair a brokenlink The connection setup delay is higher than that of table-driven protocols

(3) AODV AODV is very similar to DSR [21] It sets up aroute to the destination by sending a RouteRequest messageThe source node and the intermediate nodes store the nexthop information corresponding to each flow for data packettransmission The major difference between AODV andother reactive routing protocols is that it uses a destinationsequence number (DesSeqNum) to find the latest route tothe destination A node updates its path destination only ifthe DesSeqNum of the current packet received is greater thanthe last DesSeqNum stored at the node However AODVrequires more time to set up a connection than some otherapproaches [15 17 21]

32 Simulation Environment and Performance EvaluationThe simulator for evaluating three routing protocols is theNetwork Simulator (NS2 version 233) NS2 provides sub-stantial support for simulation of wireless networks usingdiscrete-event mode

(1) Basic Scenario The simulation scenario is designedaccording to the normal state of car running on a freewayshown in Figure 3 Assume that the freeway is 30meters wideand 500 meters long 40 vehicles are randomly distributedto the four bidirectional lanes of the freeway Each vehicleis regarded as a mobile node moving forward in a randomfashion The maximum velocity of nodes is 35ms withsimulation period of 200 seconds The channel capacity is

Table 2 Packet loss rate of three routing protocols

Time (second) DSDV () DSR () AODV ()100 1272 068 107200 762 048 048

2Mbs The MAC type is IEEE 80211 The CBR traffic modelis used with data packet size 512 bytes and sending rate160Kbps(2) Performance Evaluation The performance evaluationof routing protocols is based on the measurement of thefollowing parameters [16]Packet Loss Rate () Packet loss occurs when one or morepackets of data traveling across a VANET fail to reach theirdestinations The packet loss rate is calculated as

119901 = (1 minus119873119903

119873119904

) times 100 (1)

where 119873119903and 119873

119904represent the number of data packet

received and sent respectivelyEnd-to-End Delay (ms) End-to-end delay is defined as thetime taken for a data packet to be transmitted across thenetwork from the source to the destinationPacket Jitter (ms) Packet jitter is defined as the delay variationbetween two consecutively received packets belonging to thesame stream In general it is expressed as an absolute valueof delay variationThroughput (Kilobits per Second) System throughput is thesum of the data rates that are delivered to all nodes in thenetworkProtocol Overhead Protocol overhead refers to the number ofroutingmessages requestedwhen a data packet is successfullydelivered to the destination

33 Simulation Results and Discussion Throughout the sim-ulation we discuss the results as below

(1) Packet Loss Rate Table 2 shows the packet loss rateincurred by DSDV AODV and DSR Both DSR and AODVon-demand routing schemes have considerably less packetloss rate than DSDV Because each node is fast mobile in arandom fashion the network topology often changes Thesepackets sent may be lost once the routing table is not timelyupdated in DSDV This results in a higher packet loss rate ofDSDV

(2) End-to-End Delay Figure 4 shows the end-to-end delayof data packets The statistical values of them are (7440 plusmn

4157)ms (DSDV confidence level 120572 = 095) (8167 plusmn

9210)ms (DSR 120572 = 095) and (8107 plusmn 5026)ms (AODV120572 = 095) respectively They have no significant differencein the mean values However as a whole the margin of delayfluctuation in DSR is the highest among the three protocolsIn DSR a neighbor displacement is noticed only after apacket is sent explicitly to that node The network reacts

International Journal of Distributed Sensor Networks 5

0

001

002

003

004

005

006

007

Pack

et d

elay

(s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

0 1000 2000 3000 4000 5000 6000 70000

005

015

01

025

02

035

03

Packet ID

Pack

et d

elay

(s)

(b) DSR

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

008

Packet ID

Pack

et d

elay

(s)

(c) AODV

Figure 4 Packet delay of three routing protocols

if an acknowledgement is not received Consequently thisincreases packet delay since the packet must wait until a newroute is established(3) Packet Jitter Figure 5 shows the jitter of data packetsusing three different routing protocols The statistical valuesof them are (2356 plusmn 4199)ms (DSDV 120572 = 095) (3254 plusmn

10630)ms (DSR 120572 = 095) and (4498 plusmn 4498)ms (AODV120572 = 095) respectively DSDV shows the most superioritythan others The average jitter of AODV is the biggest amongthe three protocols Moreover the standard deviation of jitterin AODV is larger than that in DSDV Although the averagejitter of DSR is not the largest among the three protocolsDSR has the largest jitter fluctuation among themThismeansviolent variation exists in the delay of a few data packets sentin DSRThe rapid change of jitter is attributed to the frequentchange of network topology and the mechanism of inherentrouting update(4) Throughput Figure 6 shows the throughput compari-son of DSDV DSR and AODV All throughputs are ever-increasing with the time in general This can be attributed

to more active nodes to join network communication withtime extensionThe graph also reveals that AODV has higherthroughput than DSR In DSR a route is chosen basedon the short delay at the instance of route establishmentAlthough this path may be the best route at that instantit may be also a route that lacks routing stability or hasunacceptably high load In contrast AODVhas amore robustupdate mechanism to avoid bottleneck and congestion andeventually improve throughput In DSDV high packet lossrate causes throughput to drop

(5) Protocol Overhead Table 3 shows the overhead compar-ison of DSDV AODV and DSR AODV has the highestoverhead among three routing protocols due to three mainreasons Firstly AODV allows broadcast Although the dis-covery packets are broadcasted only when necessary suchas establishing a new route link breakage or route errorthe broadcast instances will often appear for a fast mobileVANET Secondly AODV allows mobile nodes to respondto link breakages and changes in network topology in a

6 International Journal of Distributed Sensor Networks

0

001

002

003

004

005

006

007

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

0

005

015

01

025

02

035

03

(b) DSR

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

Packet ID

(c) AODV

Figure 5 Packet jitter of three routing protocols

timely mannerThus numerous routingmessages are used tomaintain an active route in AODVThirdly AODV is derivedfrom DSDV It still has similar features of proactive routingprotocols When the network topology is often changedbecause of the fast mobility of nodes proactive protocolsmust send more messages to maintain a valid routing tableThe experimental results are shown in Table 3 DSR hasthe best protocol overhead performance among the threeprotocols The DSR protocol is composed of the two mainmechanisms of ldquoroute discoveryrdquo and ldquoroute maintenancerdquowhich work together to allow nodes to discover andmaintainroutes to arbitrary destinations in the ad hoc network Somemeasures reducing overhead are adopted in the process ofldquoRoute Discoveryrdquo and ldquoRoute Maintenancerdquo of DSR [21]

34 Conclusion Based on the simulation results we discov-ered that DSDV achieves marginal better packet delay andpacket jitter than AODV and DSR However it has signif-icantly higher packet loss rate and the smallest throughoutamong them Although DSR has the best protocol overhead

Table 3 Overhead of three routing protocols

Type DSDV AODV DSRNumber of received data packets 5579 5999 5993Number of sent message packets 25607 39488 1615Protocol overhead 459 658 027

performance among three routing protocols it has poorerjitter and throughput thanAODV As a whole AODV ismoreappropriate for the freeway VANET according to the qualityof service and the real time of packet delivery The inherentreason producing this result is the effect of node mobility onthe performance of the routing protocol

4 GPS Based AODV Protocol forV2V Communication [22]

As we can find through the previous protocol evaluationAODV is appropriate for the dynamic structure with mobilevehicles The source node will find a route to destination

International Journal of Distributed Sensor Networks 7

0 50 100 150 2000

20

40

60

80

100

120

140

Time (s)

Thro

ughp

ut (k

bps)

DSRDSDVAODV

Figure 6 Throughput comparison of three routing protocols

for data transmission To find a route to the destination thesource broadcasts a route request packet (RREQ) The nodereceiving RREQ will reply a route reply packet (RREP) to thereverse path that RREQ went through if it is the destinationof RREQ or if it has a recent route to the destinationOtherwise the node will broadcast RREQ until either ofthe two situations above occurs As RREP traverses backto the source the nodes along the path enter the forwardroute into their routing tables If a node leaves the networkthe node that discovers this broken link will send a linkfailure notification (RERR) to the precursorsThe RERR goesupstreamuntil it reaches the sourceThe sourcewill thereafterrestart route discovery process if needed [23]

However AODV is dedicated forMANET andmodifica-tion is necessary for VANET applications [24] Since nodes inVANET are fast moving vehicles the topology of the networkchanges all the time making the position and velocity ofthe node critical factors while finding the routes Thereforebesides conventional topology-based routing protocol (suchas AODV) position-based routing protocol which primarilyuses the position information obtained byGPS to find a routeis applied in VANET [25]

In recent years many researchers have tried to combinetwo types of routing together In [26] Kim et al pro-posed AODV-RRS which restricts the number of forward-ing RREQs according to the concepts of stable zone andcaution zone PAODV [27] restricts the number of floodingRREQs based on the distance between current node andits neighbors Both of them have similar mechanism andreduce the number of broken links DAODV [28] establishesa route depending on the direction and position of sourceintermediate and destination node Although it also reducesthe number of broken links DAODV assumes that sourcenode knows the direction and position of all the nodes inthe network Actually this is impossible with the current GPS

devices We should find some mechanisms to get the geo-graphic information of all the nodesThereby in [29] Asenovand Hnatyshin proposed GeoAODV routing protocol Eachnode will maintain an additional table geotable to keep trackof the geographic information of all the nodes RREQs arerestricted flooding in the region determined by the geotableHowever it costs more resources to maintain two tables

We go a step further in this section by proposing aGPS based AODV (GBAODV) which enhances the overallperformance of AODV in VANET In order to be attainablein physical implementation GBAODV assumes that eachnode only knows its own position and velocity Withoutmaintaining a geotable [29] GBAODV is muchmore concisethan GeoAODV

41 Overview of GBAODV With current GPS device wecan obtain the longitude (119909 coordinate) and latitude (119910coordinate) of current node and calculate the speed in eachdirection with two successive sets of coordinates Withoutmaintaining a geotable we add geographic information intorouting table to make the algorithm concise

There are two main features in GBAODV First is toreduce the number of RREQs The node receiving an RREQpacket will check the distance and motion trend betweenthe precursor and itself so as to decide if this RREQ shouldbe broadcasted By restricting these flooding RREQs it willavoid lots of packet collisions in the network As a resultthe packet delivery ratio and throughput of the networkcan be raised Second is to mark the route according to thepositions and velocities of source intermediate and desti-nation Higher mark means higher stability Consequentlyevery node will choose a route with higher mark

These two features require specifying flooding rules andmarking standards as well as some modifications in therouting table and routing packets

42 Modified Structure of Routing Table RREQ and RREP

(1) Modified Routing Table Original routing table containsthe IP addresses of next hop and destination sequencenumber of destination and total hops The routing table willupdate if the sequence number of destination is larger ortotal hops is less We added the position and velocity of thedestination node and the mark of this route into the routingtable Besides the two conditions above the routing table willupdate when the mark is larger which means that the routepath is more stable

(2) Modified Frame Structure of RREQ We inserted position(119909 119910 coordinates) and velocity (119909 119910 direction) of the sourceand current node into RREQ packet We also applied areserved segment to record the mark of the route link fromsource to precursor

(3) Modified Frame Structure of RREP RREP is modifiedsimilar to RREQ We inserted position and velocity of thedestination and current node into RREP packet We also

8 International Journal of Distributed Sensor Networks

P

B

A

C

DE

04 lowast R

08 lowast R

Figure 7 B C and D will broadcast RREQ received from P

applied a reserved segment to record the mark of the routefrom destination to current node

43 Flooding Rules and Marking Standards

(1) Flooding Rules Let 119889 stand for the distance between theprecursor and current node and 119877 stand for the maximumtransmission distance of the wireless cardThe flooding rulesare as follows

If (119889 lt 04 119877 (119889 gt 08119877ampamp119889 is increasing))

discard RREQ

Else

broadcast RREQ

The coefficients 04 and 08 are chosen empiricallyWe conducted simulations with the combinations of thesecoefficients to study numbers of RREQs and RERRs packetloss ratio and average end-to-end delay Results show that thecombination of 04 and 08 is the best among all

In Figure 7 B C andDwill broadcast the RREQ receivedfrom P while A and E will discard it

(2) Marking Standards As illustrated in Figure 8 let 119871119863119888

stand for the current distance between node 119871 and node 119868119871119863119897stand for this distance a moment later 119877119863

119888stand for the

current distance between node 119868 and node 119877 and 119877119863119897stand

for this distance a moment later We specify the markingstandards as follows

mark = 2 sgn (119871119863119888minus 119871119863119897) + 2 sgn (119877119863

119888minus 119877119863119897) + 119862 (2)

where sgn() is a signum function119862 is a constant

119862 = 0 if 119868 is in the dashed rectangleminus3 if 119868 is out of this rectangle

(3)

minus3 is chosen experimentally according to simulationsSince the transmission distance of wireless LAN card is

limited geographic distance of each hop is the key factor thataffects the quality of communication If the distance stays

L

R

I1

I2

Figure 8 Different mark for different regions

unchanged then the route is really stable so that the sourceand destination can communicate steadily If the distance isincreasing it is possible that one node in this routewill exceedthe transmission distance soon If the distance is decreasingthen after some time the two nodes may cross over and startto depart It is better than increasing distance but not asgood as unchanged distance So we set two values for eachchanged distance 0 for each decreasing distance and minus2 foreach increasing distance as shown in (2)

To find a path to the destination the source broadcastsan RREQ containing the position and velocity of source andpreset a standard mark value

As illustrated in Figure 9 when a node receives an RREQthe following will occur

(i) If it does not have a route to the source node it willinsert the route RREQ into its routing table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination or has a route to destination itwill create an RREP in which the mark is a standardvalue (if it is the destination) or is obtained fromrouting table (if it has a route to destination)

(iv) If it is a node besides (3) and satisfies the broadcastingconditions (refer to flooding rules) it will add themark of sourcerarr precursorrarr current to the presentmark in RREQ It will also update the position andvelocity of current node in RREQ

As illustrated in Figure 10 when a node receives an RREPthe following will occur

(i) If it does not have a route to the destination it willinsert the route which RREP goes through into itsrouting table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination of RREP (ie the source ofRREQ) it will insert the route into routing table

(iv) If it is a forwarding node before unicasting RREQit will add the mark of sourcerarr precursorrarr current

International Journal of Distributed Sensor Networks 9

Receives anRREQ

Has a route tosource

Inserts inrouting table

NeedsupdateUpdates

Destination or has aroute to destination

Replies anRREP

Needsforward

DiscardsRREQ

Updates markand current

nodersquos positionand velocity in

RREQ

BroadcastsRREQ

No

Yes

Yes

No

YesNo

No

Yes

Figure 9 Flowchart of processing RREQ

to the present mark in RREP It will also update theposition and velocity of current node in RREQ

44 Simulation Setup

(1) Simulation Tools We chose VanetMobiSim 11 [30] fortraffic simulation This software can generate a traffic flow inthe format suitable forNS2 [31] which can loadGBAODV fornetwork simulation

With VanetMobiSim we import maps from the USCensus Bureau TIGERLine database [32] which includescomplete coverage of the United States Puerto Rico and soforthMoreoverVanetMobiSim supports formultilane roadsdifferentiated speed constraints and traffic light signals atintersections All the vehicles can be set to Intelligent DriverModel with Lane Changing (IDM LC) [33 34] For thesereasons the scenario in traffic layer is quite authentic whichmakes the simulation in network layer reliable

(2) Parameter Settings In our simulation we observed twotypes of traffic models downtown and highway TGR11001[32] (district of Columbia WA) is chosen as downtown

Receives anRREP

Has a route todestination

Inserts inrouting table

NeedsupdateUpdates

Inserts inrouting table

Sourcenode

Update mark andcurrent nodersquosposition and

velocity in RREP

Unicast RREP

NoYes

Yes

No

Yes

No

Figure 10 Flowchart of processing RREP

Table 4 Parameter settings of traffic

Traffic layer Downtown HighwayArea (m2) 1000 times 1000 1000 times 1000

Number of lanes 3 3Maximum number of traffic lights 10 NoneSpeed (kmh) 20sim80 60sim120Simulation time (s) 250 1000

Table 5 Parameter settings of network

Network layer Downtown HighwayMaximum transmission distance (m) 250 250Number of sources 35 17Number of connections 56 26CBR packet size (bytes) 256 512Transmission rate (pkts) 1 2Simulation time (s) 250 1000

map and TGR36001 [32] (Albany county NY) is chosen ashighway map Multilanes and traffic lights are involved Allthe vehicles follow the IDM LC driving model Tables 4 and5 are the parameter settings of traffic and network simulationTraffic flow and CBR (constant bitrate) data flow are bothgenerated randomly

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 3: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of Distributed Sensor Networks 3

Table 1 Sensors for vehicular information acquistion

Sensor FunctionAirbag deployment LED(on dashboard) Airbag deployment detection

Vehicle wheel sensor Detect vehicle speedGPSSiRF starIII Vehicle localization6G one-axis accelerometer Sudden slow down detection500G one-axisaccelerometer Slight bumping detection

15 G three-axisaccelerometer Rolling over

Steering wheel angularsensor S-shape steering detection

Slight bumping2

15G accelerometer

reading abnormal500G accelerometer

reading abnormal6G accelerometer

reading abnormal

Airbag deployed

Sudden slowdown1

Airbag deployed3

Level =

Level =

Level =

Level =

Airbag deployed4

Level = 0

Yes

Yes

Yes

Yes

No

No

No

No

Figure 1 Flowchart for collision estimation

3 Evaluation of Classical RoutingProtocols in MANET [13]

The speed of vehicle moving on a freeway is fairly fast atapproximately 80 kmh in general When the accident occurson a freeway the warning message must be instantly sentto other vehicles nearby via the VANET so that the driversnearby have enough time to avoid collision In order toachieve reliable real-time communication the performanceof routing protocol used by the VANET is important [5 14]Different routing protocols have different network charac-teristics [15ndash17] VANET on the freeway is a fast and highlydynamicmobile ad hocwireless networkwithout fixed routerhosts or wireless based stations A significant challenge in thefreeway VANET is how to select an efficient routing protocolamong numerous protocols Thus three classical routing

Real-time vehicle statemonitoring

Real-time estimation ofcollision

Is vehiclecondition safe

Severity level = 1

Report to traffic centerand forward to nearby

vehicles

Forward waning tonearby vehicles

through VANET

Yes

Yes

No

No

Figure 2 Algorithm flowchart of the collision evaluation

protocols are evaluated and compared A proper routingprotocol will be selected based on which modification canbe made to bear the unique features for VANET

31 Classical Routing Protocols for MANET Numerous adhoc routing protocols have been proposed by the InternetEngineering Task Force (IETF) Mobile Ad hoc Networks(MANET) Working Group [18] These routing protocols aredivided into two major classes based on the underlyingrouting information update mechanism employed reactive(on-demand) or proactive (table-driven) For performanceevaluation we have compared three typical routing protocolsin MANET for V2V application Destination sequenceddistance vector protocol (DSDV) is selected as an example fortable-driven protocols while both dynamic source routing(DSR) and ad hoc on-demand distance vector (AODV)protocols are selected as examples for on-demand protocols[19ndash21]

(1) DSDV DSDV is based on the Bellman-Ford algorithmwhich can effectively solve routing loop problem [19] Eachnode has a routing table which contains the shortest path toevery other node in the network Each entry in the routingtable contains a sequence number The number is generatedby the destination If a node receives new informationit consults the routing table and uses the latest sequencenumber to forward If the sequence number is the same asthe one existing in the table the route with the bettermetric isused Stale entries are deleted by regular update of its routingtables DSDV is suitable for creating a small-scale ad hocnetwork

(2) DSR DSR uses source routing instead of hop-by-hoppacket routing in comparison with DSDV and has two majorphases which are route discovery and route maintenance[20] Route discovery is used to set up a route from sourcenode to destination by sending RouteRequest packet in

4 International Journal of Distributed Sensor Networks

Figure 3 Simulation scenario of vehicles moving on a freeway

the source node If a node in the path moves away and breakswireless communication route maintenance will rebuilda route from source node to destination one by sendingRouteError packet to the node adjacent to broken link Eachdata packet carries corresponding routing informationThusit eliminates the need to periodically flood the network withtable update messages which are required in a table-drivenapproach However DSR does not locally repair a brokenlink The connection setup delay is higher than that of table-driven protocols

(3) AODV AODV is very similar to DSR [21] It sets up aroute to the destination by sending a RouteRequest messageThe source node and the intermediate nodes store the nexthop information corresponding to each flow for data packettransmission The major difference between AODV andother reactive routing protocols is that it uses a destinationsequence number (DesSeqNum) to find the latest route tothe destination A node updates its path destination only ifthe DesSeqNum of the current packet received is greater thanthe last DesSeqNum stored at the node However AODVrequires more time to set up a connection than some otherapproaches [15 17 21]

32 Simulation Environment and Performance EvaluationThe simulator for evaluating three routing protocols is theNetwork Simulator (NS2 version 233) NS2 provides sub-stantial support for simulation of wireless networks usingdiscrete-event mode

(1) Basic Scenario The simulation scenario is designedaccording to the normal state of car running on a freewayshown in Figure 3 Assume that the freeway is 30meters wideand 500 meters long 40 vehicles are randomly distributedto the four bidirectional lanes of the freeway Each vehicleis regarded as a mobile node moving forward in a randomfashion The maximum velocity of nodes is 35ms withsimulation period of 200 seconds The channel capacity is

Table 2 Packet loss rate of three routing protocols

Time (second) DSDV () DSR () AODV ()100 1272 068 107200 762 048 048

2Mbs The MAC type is IEEE 80211 The CBR traffic modelis used with data packet size 512 bytes and sending rate160Kbps(2) Performance Evaluation The performance evaluationof routing protocols is based on the measurement of thefollowing parameters [16]Packet Loss Rate () Packet loss occurs when one or morepackets of data traveling across a VANET fail to reach theirdestinations The packet loss rate is calculated as

119901 = (1 minus119873119903

119873119904

) times 100 (1)

where 119873119903and 119873

119904represent the number of data packet

received and sent respectivelyEnd-to-End Delay (ms) End-to-end delay is defined as thetime taken for a data packet to be transmitted across thenetwork from the source to the destinationPacket Jitter (ms) Packet jitter is defined as the delay variationbetween two consecutively received packets belonging to thesame stream In general it is expressed as an absolute valueof delay variationThroughput (Kilobits per Second) System throughput is thesum of the data rates that are delivered to all nodes in thenetworkProtocol Overhead Protocol overhead refers to the number ofroutingmessages requestedwhen a data packet is successfullydelivered to the destination

33 Simulation Results and Discussion Throughout the sim-ulation we discuss the results as below

(1) Packet Loss Rate Table 2 shows the packet loss rateincurred by DSDV AODV and DSR Both DSR and AODVon-demand routing schemes have considerably less packetloss rate than DSDV Because each node is fast mobile in arandom fashion the network topology often changes Thesepackets sent may be lost once the routing table is not timelyupdated in DSDV This results in a higher packet loss rate ofDSDV

(2) End-to-End Delay Figure 4 shows the end-to-end delayof data packets The statistical values of them are (7440 plusmn

4157)ms (DSDV confidence level 120572 = 095) (8167 plusmn

9210)ms (DSR 120572 = 095) and (8107 plusmn 5026)ms (AODV120572 = 095) respectively They have no significant differencein the mean values However as a whole the margin of delayfluctuation in DSR is the highest among the three protocolsIn DSR a neighbor displacement is noticed only after apacket is sent explicitly to that node The network reacts

International Journal of Distributed Sensor Networks 5

0

001

002

003

004

005

006

007

Pack

et d

elay

(s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

0 1000 2000 3000 4000 5000 6000 70000

005

015

01

025

02

035

03

Packet ID

Pack

et d

elay

(s)

(b) DSR

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

008

Packet ID

Pack

et d

elay

(s)

(c) AODV

Figure 4 Packet delay of three routing protocols

if an acknowledgement is not received Consequently thisincreases packet delay since the packet must wait until a newroute is established(3) Packet Jitter Figure 5 shows the jitter of data packetsusing three different routing protocols The statistical valuesof them are (2356 plusmn 4199)ms (DSDV 120572 = 095) (3254 plusmn

10630)ms (DSR 120572 = 095) and (4498 plusmn 4498)ms (AODV120572 = 095) respectively DSDV shows the most superioritythan others The average jitter of AODV is the biggest amongthe three protocols Moreover the standard deviation of jitterin AODV is larger than that in DSDV Although the averagejitter of DSR is not the largest among the three protocolsDSR has the largest jitter fluctuation among themThismeansviolent variation exists in the delay of a few data packets sentin DSRThe rapid change of jitter is attributed to the frequentchange of network topology and the mechanism of inherentrouting update(4) Throughput Figure 6 shows the throughput compari-son of DSDV DSR and AODV All throughputs are ever-increasing with the time in general This can be attributed

to more active nodes to join network communication withtime extensionThe graph also reveals that AODV has higherthroughput than DSR In DSR a route is chosen basedon the short delay at the instance of route establishmentAlthough this path may be the best route at that instantit may be also a route that lacks routing stability or hasunacceptably high load In contrast AODVhas amore robustupdate mechanism to avoid bottleneck and congestion andeventually improve throughput In DSDV high packet lossrate causes throughput to drop

(5) Protocol Overhead Table 3 shows the overhead compar-ison of DSDV AODV and DSR AODV has the highestoverhead among three routing protocols due to three mainreasons Firstly AODV allows broadcast Although the dis-covery packets are broadcasted only when necessary suchas establishing a new route link breakage or route errorthe broadcast instances will often appear for a fast mobileVANET Secondly AODV allows mobile nodes to respondto link breakages and changes in network topology in a

6 International Journal of Distributed Sensor Networks

0

001

002

003

004

005

006

007

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

0

005

015

01

025

02

035

03

(b) DSR

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

Packet ID

(c) AODV

Figure 5 Packet jitter of three routing protocols

timely mannerThus numerous routingmessages are used tomaintain an active route in AODVThirdly AODV is derivedfrom DSDV It still has similar features of proactive routingprotocols When the network topology is often changedbecause of the fast mobility of nodes proactive protocolsmust send more messages to maintain a valid routing tableThe experimental results are shown in Table 3 DSR hasthe best protocol overhead performance among the threeprotocols The DSR protocol is composed of the two mainmechanisms of ldquoroute discoveryrdquo and ldquoroute maintenancerdquowhich work together to allow nodes to discover andmaintainroutes to arbitrary destinations in the ad hoc network Somemeasures reducing overhead are adopted in the process ofldquoRoute Discoveryrdquo and ldquoRoute Maintenancerdquo of DSR [21]

34 Conclusion Based on the simulation results we discov-ered that DSDV achieves marginal better packet delay andpacket jitter than AODV and DSR However it has signif-icantly higher packet loss rate and the smallest throughoutamong them Although DSR has the best protocol overhead

Table 3 Overhead of three routing protocols

Type DSDV AODV DSRNumber of received data packets 5579 5999 5993Number of sent message packets 25607 39488 1615Protocol overhead 459 658 027

performance among three routing protocols it has poorerjitter and throughput thanAODV As a whole AODV ismoreappropriate for the freeway VANET according to the qualityof service and the real time of packet delivery The inherentreason producing this result is the effect of node mobility onthe performance of the routing protocol

4 GPS Based AODV Protocol forV2V Communication [22]

As we can find through the previous protocol evaluationAODV is appropriate for the dynamic structure with mobilevehicles The source node will find a route to destination

International Journal of Distributed Sensor Networks 7

0 50 100 150 2000

20

40

60

80

100

120

140

Time (s)

Thro

ughp

ut (k

bps)

DSRDSDVAODV

Figure 6 Throughput comparison of three routing protocols

for data transmission To find a route to the destination thesource broadcasts a route request packet (RREQ) The nodereceiving RREQ will reply a route reply packet (RREP) to thereverse path that RREQ went through if it is the destinationof RREQ or if it has a recent route to the destinationOtherwise the node will broadcast RREQ until either ofthe two situations above occurs As RREP traverses backto the source the nodes along the path enter the forwardroute into their routing tables If a node leaves the networkthe node that discovers this broken link will send a linkfailure notification (RERR) to the precursorsThe RERR goesupstreamuntil it reaches the sourceThe sourcewill thereafterrestart route discovery process if needed [23]

However AODV is dedicated forMANET andmodifica-tion is necessary for VANET applications [24] Since nodes inVANET are fast moving vehicles the topology of the networkchanges all the time making the position and velocity ofthe node critical factors while finding the routes Thereforebesides conventional topology-based routing protocol (suchas AODV) position-based routing protocol which primarilyuses the position information obtained byGPS to find a routeis applied in VANET [25]

In recent years many researchers have tried to combinetwo types of routing together In [26] Kim et al pro-posed AODV-RRS which restricts the number of forward-ing RREQs according to the concepts of stable zone andcaution zone PAODV [27] restricts the number of floodingRREQs based on the distance between current node andits neighbors Both of them have similar mechanism andreduce the number of broken links DAODV [28] establishesa route depending on the direction and position of sourceintermediate and destination node Although it also reducesthe number of broken links DAODV assumes that sourcenode knows the direction and position of all the nodes inthe network Actually this is impossible with the current GPS

devices We should find some mechanisms to get the geo-graphic information of all the nodesThereby in [29] Asenovand Hnatyshin proposed GeoAODV routing protocol Eachnode will maintain an additional table geotable to keep trackof the geographic information of all the nodes RREQs arerestricted flooding in the region determined by the geotableHowever it costs more resources to maintain two tables

We go a step further in this section by proposing aGPS based AODV (GBAODV) which enhances the overallperformance of AODV in VANET In order to be attainablein physical implementation GBAODV assumes that eachnode only knows its own position and velocity Withoutmaintaining a geotable [29] GBAODV is muchmore concisethan GeoAODV

41 Overview of GBAODV With current GPS device wecan obtain the longitude (119909 coordinate) and latitude (119910coordinate) of current node and calculate the speed in eachdirection with two successive sets of coordinates Withoutmaintaining a geotable we add geographic information intorouting table to make the algorithm concise

There are two main features in GBAODV First is toreduce the number of RREQs The node receiving an RREQpacket will check the distance and motion trend betweenthe precursor and itself so as to decide if this RREQ shouldbe broadcasted By restricting these flooding RREQs it willavoid lots of packet collisions in the network As a resultthe packet delivery ratio and throughput of the networkcan be raised Second is to mark the route according to thepositions and velocities of source intermediate and desti-nation Higher mark means higher stability Consequentlyevery node will choose a route with higher mark

These two features require specifying flooding rules andmarking standards as well as some modifications in therouting table and routing packets

42 Modified Structure of Routing Table RREQ and RREP

(1) Modified Routing Table Original routing table containsthe IP addresses of next hop and destination sequencenumber of destination and total hops The routing table willupdate if the sequence number of destination is larger ortotal hops is less We added the position and velocity of thedestination node and the mark of this route into the routingtable Besides the two conditions above the routing table willupdate when the mark is larger which means that the routepath is more stable

(2) Modified Frame Structure of RREQ We inserted position(119909 119910 coordinates) and velocity (119909 119910 direction) of the sourceand current node into RREQ packet We also applied areserved segment to record the mark of the route link fromsource to precursor

(3) Modified Frame Structure of RREP RREP is modifiedsimilar to RREQ We inserted position and velocity of thedestination and current node into RREP packet We also

8 International Journal of Distributed Sensor Networks

P

B

A

C

DE

04 lowast R

08 lowast R

Figure 7 B C and D will broadcast RREQ received from P

applied a reserved segment to record the mark of the routefrom destination to current node

43 Flooding Rules and Marking Standards

(1) Flooding Rules Let 119889 stand for the distance between theprecursor and current node and 119877 stand for the maximumtransmission distance of the wireless cardThe flooding rulesare as follows

If (119889 lt 04 119877 (119889 gt 08119877ampamp119889 is increasing))

discard RREQ

Else

broadcast RREQ

The coefficients 04 and 08 are chosen empiricallyWe conducted simulations with the combinations of thesecoefficients to study numbers of RREQs and RERRs packetloss ratio and average end-to-end delay Results show that thecombination of 04 and 08 is the best among all

In Figure 7 B C andDwill broadcast the RREQ receivedfrom P while A and E will discard it

(2) Marking Standards As illustrated in Figure 8 let 119871119863119888

stand for the current distance between node 119871 and node 119868119871119863119897stand for this distance a moment later 119877119863

119888stand for the

current distance between node 119868 and node 119877 and 119877119863119897stand

for this distance a moment later We specify the markingstandards as follows

mark = 2 sgn (119871119863119888minus 119871119863119897) + 2 sgn (119877119863

119888minus 119877119863119897) + 119862 (2)

where sgn() is a signum function119862 is a constant

119862 = 0 if 119868 is in the dashed rectangleminus3 if 119868 is out of this rectangle

(3)

minus3 is chosen experimentally according to simulationsSince the transmission distance of wireless LAN card is

limited geographic distance of each hop is the key factor thataffects the quality of communication If the distance stays

L

R

I1

I2

Figure 8 Different mark for different regions

unchanged then the route is really stable so that the sourceand destination can communicate steadily If the distance isincreasing it is possible that one node in this routewill exceedthe transmission distance soon If the distance is decreasingthen after some time the two nodes may cross over and startto depart It is better than increasing distance but not asgood as unchanged distance So we set two values for eachchanged distance 0 for each decreasing distance and minus2 foreach increasing distance as shown in (2)

To find a path to the destination the source broadcastsan RREQ containing the position and velocity of source andpreset a standard mark value

As illustrated in Figure 9 when a node receives an RREQthe following will occur

(i) If it does not have a route to the source node it willinsert the route RREQ into its routing table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination or has a route to destination itwill create an RREP in which the mark is a standardvalue (if it is the destination) or is obtained fromrouting table (if it has a route to destination)

(iv) If it is a node besides (3) and satisfies the broadcastingconditions (refer to flooding rules) it will add themark of sourcerarr precursorrarr current to the presentmark in RREQ It will also update the position andvelocity of current node in RREQ

As illustrated in Figure 10 when a node receives an RREPthe following will occur

(i) If it does not have a route to the destination it willinsert the route which RREP goes through into itsrouting table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination of RREP (ie the source ofRREQ) it will insert the route into routing table

(iv) If it is a forwarding node before unicasting RREQit will add the mark of sourcerarr precursorrarr current

International Journal of Distributed Sensor Networks 9

Receives anRREQ

Has a route tosource

Inserts inrouting table

NeedsupdateUpdates

Destination or has aroute to destination

Replies anRREP

Needsforward

DiscardsRREQ

Updates markand current

nodersquos positionand velocity in

RREQ

BroadcastsRREQ

No

Yes

Yes

No

YesNo

No

Yes

Figure 9 Flowchart of processing RREQ

to the present mark in RREP It will also update theposition and velocity of current node in RREQ

44 Simulation Setup

(1) Simulation Tools We chose VanetMobiSim 11 [30] fortraffic simulation This software can generate a traffic flow inthe format suitable forNS2 [31] which can loadGBAODV fornetwork simulation

With VanetMobiSim we import maps from the USCensus Bureau TIGERLine database [32] which includescomplete coverage of the United States Puerto Rico and soforthMoreoverVanetMobiSim supports formultilane roadsdifferentiated speed constraints and traffic light signals atintersections All the vehicles can be set to Intelligent DriverModel with Lane Changing (IDM LC) [33 34] For thesereasons the scenario in traffic layer is quite authentic whichmakes the simulation in network layer reliable

(2) Parameter Settings In our simulation we observed twotypes of traffic models downtown and highway TGR11001[32] (district of Columbia WA) is chosen as downtown

Receives anRREP

Has a route todestination

Inserts inrouting table

NeedsupdateUpdates

Inserts inrouting table

Sourcenode

Update mark andcurrent nodersquosposition and

velocity in RREP

Unicast RREP

NoYes

Yes

No

Yes

No

Figure 10 Flowchart of processing RREP

Table 4 Parameter settings of traffic

Traffic layer Downtown HighwayArea (m2) 1000 times 1000 1000 times 1000

Number of lanes 3 3Maximum number of traffic lights 10 NoneSpeed (kmh) 20sim80 60sim120Simulation time (s) 250 1000

Table 5 Parameter settings of network

Network layer Downtown HighwayMaximum transmission distance (m) 250 250Number of sources 35 17Number of connections 56 26CBR packet size (bytes) 256 512Transmission rate (pkts) 1 2Simulation time (s) 250 1000

map and TGR36001 [32] (Albany county NY) is chosen ashighway map Multilanes and traffic lights are involved Allthe vehicles follow the IDM LC driving model Tables 4 and5 are the parameter settings of traffic and network simulationTraffic flow and CBR (constant bitrate) data flow are bothgenerated randomly

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 4: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

4 International Journal of Distributed Sensor Networks

Figure 3 Simulation scenario of vehicles moving on a freeway

the source node If a node in the path moves away and breakswireless communication route maintenance will rebuilda route from source node to destination one by sendingRouteError packet to the node adjacent to broken link Eachdata packet carries corresponding routing informationThusit eliminates the need to periodically flood the network withtable update messages which are required in a table-drivenapproach However DSR does not locally repair a brokenlink The connection setup delay is higher than that of table-driven protocols

(3) AODV AODV is very similar to DSR [21] It sets up aroute to the destination by sending a RouteRequest messageThe source node and the intermediate nodes store the nexthop information corresponding to each flow for data packettransmission The major difference between AODV andother reactive routing protocols is that it uses a destinationsequence number (DesSeqNum) to find the latest route tothe destination A node updates its path destination only ifthe DesSeqNum of the current packet received is greater thanthe last DesSeqNum stored at the node However AODVrequires more time to set up a connection than some otherapproaches [15 17 21]

32 Simulation Environment and Performance EvaluationThe simulator for evaluating three routing protocols is theNetwork Simulator (NS2 version 233) NS2 provides sub-stantial support for simulation of wireless networks usingdiscrete-event mode

(1) Basic Scenario The simulation scenario is designedaccording to the normal state of car running on a freewayshown in Figure 3 Assume that the freeway is 30meters wideand 500 meters long 40 vehicles are randomly distributedto the four bidirectional lanes of the freeway Each vehicleis regarded as a mobile node moving forward in a randomfashion The maximum velocity of nodes is 35ms withsimulation period of 200 seconds The channel capacity is

Table 2 Packet loss rate of three routing protocols

Time (second) DSDV () DSR () AODV ()100 1272 068 107200 762 048 048

2Mbs The MAC type is IEEE 80211 The CBR traffic modelis used with data packet size 512 bytes and sending rate160Kbps(2) Performance Evaluation The performance evaluationof routing protocols is based on the measurement of thefollowing parameters [16]Packet Loss Rate () Packet loss occurs when one or morepackets of data traveling across a VANET fail to reach theirdestinations The packet loss rate is calculated as

119901 = (1 minus119873119903

119873119904

) times 100 (1)

where 119873119903and 119873

119904represent the number of data packet

received and sent respectivelyEnd-to-End Delay (ms) End-to-end delay is defined as thetime taken for a data packet to be transmitted across thenetwork from the source to the destinationPacket Jitter (ms) Packet jitter is defined as the delay variationbetween two consecutively received packets belonging to thesame stream In general it is expressed as an absolute valueof delay variationThroughput (Kilobits per Second) System throughput is thesum of the data rates that are delivered to all nodes in thenetworkProtocol Overhead Protocol overhead refers to the number ofroutingmessages requestedwhen a data packet is successfullydelivered to the destination

33 Simulation Results and Discussion Throughout the sim-ulation we discuss the results as below

(1) Packet Loss Rate Table 2 shows the packet loss rateincurred by DSDV AODV and DSR Both DSR and AODVon-demand routing schemes have considerably less packetloss rate than DSDV Because each node is fast mobile in arandom fashion the network topology often changes Thesepackets sent may be lost once the routing table is not timelyupdated in DSDV This results in a higher packet loss rate ofDSDV

(2) End-to-End Delay Figure 4 shows the end-to-end delayof data packets The statistical values of them are (7440 plusmn

4157)ms (DSDV confidence level 120572 = 095) (8167 plusmn

9210)ms (DSR 120572 = 095) and (8107 plusmn 5026)ms (AODV120572 = 095) respectively They have no significant differencein the mean values However as a whole the margin of delayfluctuation in DSR is the highest among the three protocolsIn DSR a neighbor displacement is noticed only after apacket is sent explicitly to that node The network reacts

International Journal of Distributed Sensor Networks 5

0

001

002

003

004

005

006

007

Pack

et d

elay

(s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

0 1000 2000 3000 4000 5000 6000 70000

005

015

01

025

02

035

03

Packet ID

Pack

et d

elay

(s)

(b) DSR

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

008

Packet ID

Pack

et d

elay

(s)

(c) AODV

Figure 4 Packet delay of three routing protocols

if an acknowledgement is not received Consequently thisincreases packet delay since the packet must wait until a newroute is established(3) Packet Jitter Figure 5 shows the jitter of data packetsusing three different routing protocols The statistical valuesof them are (2356 plusmn 4199)ms (DSDV 120572 = 095) (3254 plusmn

10630)ms (DSR 120572 = 095) and (4498 plusmn 4498)ms (AODV120572 = 095) respectively DSDV shows the most superioritythan others The average jitter of AODV is the biggest amongthe three protocols Moreover the standard deviation of jitterin AODV is larger than that in DSDV Although the averagejitter of DSR is not the largest among the three protocolsDSR has the largest jitter fluctuation among themThismeansviolent variation exists in the delay of a few data packets sentin DSRThe rapid change of jitter is attributed to the frequentchange of network topology and the mechanism of inherentrouting update(4) Throughput Figure 6 shows the throughput compari-son of DSDV DSR and AODV All throughputs are ever-increasing with the time in general This can be attributed

to more active nodes to join network communication withtime extensionThe graph also reveals that AODV has higherthroughput than DSR In DSR a route is chosen basedon the short delay at the instance of route establishmentAlthough this path may be the best route at that instantit may be also a route that lacks routing stability or hasunacceptably high load In contrast AODVhas amore robustupdate mechanism to avoid bottleneck and congestion andeventually improve throughput In DSDV high packet lossrate causes throughput to drop

(5) Protocol Overhead Table 3 shows the overhead compar-ison of DSDV AODV and DSR AODV has the highestoverhead among three routing protocols due to three mainreasons Firstly AODV allows broadcast Although the dis-covery packets are broadcasted only when necessary suchas establishing a new route link breakage or route errorthe broadcast instances will often appear for a fast mobileVANET Secondly AODV allows mobile nodes to respondto link breakages and changes in network topology in a

6 International Journal of Distributed Sensor Networks

0

001

002

003

004

005

006

007

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

0

005

015

01

025

02

035

03

(b) DSR

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

Packet ID

(c) AODV

Figure 5 Packet jitter of three routing protocols

timely mannerThus numerous routingmessages are used tomaintain an active route in AODVThirdly AODV is derivedfrom DSDV It still has similar features of proactive routingprotocols When the network topology is often changedbecause of the fast mobility of nodes proactive protocolsmust send more messages to maintain a valid routing tableThe experimental results are shown in Table 3 DSR hasthe best protocol overhead performance among the threeprotocols The DSR protocol is composed of the two mainmechanisms of ldquoroute discoveryrdquo and ldquoroute maintenancerdquowhich work together to allow nodes to discover andmaintainroutes to arbitrary destinations in the ad hoc network Somemeasures reducing overhead are adopted in the process ofldquoRoute Discoveryrdquo and ldquoRoute Maintenancerdquo of DSR [21]

34 Conclusion Based on the simulation results we discov-ered that DSDV achieves marginal better packet delay andpacket jitter than AODV and DSR However it has signif-icantly higher packet loss rate and the smallest throughoutamong them Although DSR has the best protocol overhead

Table 3 Overhead of three routing protocols

Type DSDV AODV DSRNumber of received data packets 5579 5999 5993Number of sent message packets 25607 39488 1615Protocol overhead 459 658 027

performance among three routing protocols it has poorerjitter and throughput thanAODV As a whole AODV ismoreappropriate for the freeway VANET according to the qualityof service and the real time of packet delivery The inherentreason producing this result is the effect of node mobility onthe performance of the routing protocol

4 GPS Based AODV Protocol forV2V Communication [22]

As we can find through the previous protocol evaluationAODV is appropriate for the dynamic structure with mobilevehicles The source node will find a route to destination

International Journal of Distributed Sensor Networks 7

0 50 100 150 2000

20

40

60

80

100

120

140

Time (s)

Thro

ughp

ut (k

bps)

DSRDSDVAODV

Figure 6 Throughput comparison of three routing protocols

for data transmission To find a route to the destination thesource broadcasts a route request packet (RREQ) The nodereceiving RREQ will reply a route reply packet (RREP) to thereverse path that RREQ went through if it is the destinationof RREQ or if it has a recent route to the destinationOtherwise the node will broadcast RREQ until either ofthe two situations above occurs As RREP traverses backto the source the nodes along the path enter the forwardroute into their routing tables If a node leaves the networkthe node that discovers this broken link will send a linkfailure notification (RERR) to the precursorsThe RERR goesupstreamuntil it reaches the sourceThe sourcewill thereafterrestart route discovery process if needed [23]

However AODV is dedicated forMANET andmodifica-tion is necessary for VANET applications [24] Since nodes inVANET are fast moving vehicles the topology of the networkchanges all the time making the position and velocity ofthe node critical factors while finding the routes Thereforebesides conventional topology-based routing protocol (suchas AODV) position-based routing protocol which primarilyuses the position information obtained byGPS to find a routeis applied in VANET [25]

In recent years many researchers have tried to combinetwo types of routing together In [26] Kim et al pro-posed AODV-RRS which restricts the number of forward-ing RREQs according to the concepts of stable zone andcaution zone PAODV [27] restricts the number of floodingRREQs based on the distance between current node andits neighbors Both of them have similar mechanism andreduce the number of broken links DAODV [28] establishesa route depending on the direction and position of sourceintermediate and destination node Although it also reducesthe number of broken links DAODV assumes that sourcenode knows the direction and position of all the nodes inthe network Actually this is impossible with the current GPS

devices We should find some mechanisms to get the geo-graphic information of all the nodesThereby in [29] Asenovand Hnatyshin proposed GeoAODV routing protocol Eachnode will maintain an additional table geotable to keep trackof the geographic information of all the nodes RREQs arerestricted flooding in the region determined by the geotableHowever it costs more resources to maintain two tables

We go a step further in this section by proposing aGPS based AODV (GBAODV) which enhances the overallperformance of AODV in VANET In order to be attainablein physical implementation GBAODV assumes that eachnode only knows its own position and velocity Withoutmaintaining a geotable [29] GBAODV is muchmore concisethan GeoAODV

41 Overview of GBAODV With current GPS device wecan obtain the longitude (119909 coordinate) and latitude (119910coordinate) of current node and calculate the speed in eachdirection with two successive sets of coordinates Withoutmaintaining a geotable we add geographic information intorouting table to make the algorithm concise

There are two main features in GBAODV First is toreduce the number of RREQs The node receiving an RREQpacket will check the distance and motion trend betweenthe precursor and itself so as to decide if this RREQ shouldbe broadcasted By restricting these flooding RREQs it willavoid lots of packet collisions in the network As a resultthe packet delivery ratio and throughput of the networkcan be raised Second is to mark the route according to thepositions and velocities of source intermediate and desti-nation Higher mark means higher stability Consequentlyevery node will choose a route with higher mark

These two features require specifying flooding rules andmarking standards as well as some modifications in therouting table and routing packets

42 Modified Structure of Routing Table RREQ and RREP

(1) Modified Routing Table Original routing table containsthe IP addresses of next hop and destination sequencenumber of destination and total hops The routing table willupdate if the sequence number of destination is larger ortotal hops is less We added the position and velocity of thedestination node and the mark of this route into the routingtable Besides the two conditions above the routing table willupdate when the mark is larger which means that the routepath is more stable

(2) Modified Frame Structure of RREQ We inserted position(119909 119910 coordinates) and velocity (119909 119910 direction) of the sourceand current node into RREQ packet We also applied areserved segment to record the mark of the route link fromsource to precursor

(3) Modified Frame Structure of RREP RREP is modifiedsimilar to RREQ We inserted position and velocity of thedestination and current node into RREP packet We also

8 International Journal of Distributed Sensor Networks

P

B

A

C

DE

04 lowast R

08 lowast R

Figure 7 B C and D will broadcast RREQ received from P

applied a reserved segment to record the mark of the routefrom destination to current node

43 Flooding Rules and Marking Standards

(1) Flooding Rules Let 119889 stand for the distance between theprecursor and current node and 119877 stand for the maximumtransmission distance of the wireless cardThe flooding rulesare as follows

If (119889 lt 04 119877 (119889 gt 08119877ampamp119889 is increasing))

discard RREQ

Else

broadcast RREQ

The coefficients 04 and 08 are chosen empiricallyWe conducted simulations with the combinations of thesecoefficients to study numbers of RREQs and RERRs packetloss ratio and average end-to-end delay Results show that thecombination of 04 and 08 is the best among all

In Figure 7 B C andDwill broadcast the RREQ receivedfrom P while A and E will discard it

(2) Marking Standards As illustrated in Figure 8 let 119871119863119888

stand for the current distance between node 119871 and node 119868119871119863119897stand for this distance a moment later 119877119863

119888stand for the

current distance between node 119868 and node 119877 and 119877119863119897stand

for this distance a moment later We specify the markingstandards as follows

mark = 2 sgn (119871119863119888minus 119871119863119897) + 2 sgn (119877119863

119888minus 119877119863119897) + 119862 (2)

where sgn() is a signum function119862 is a constant

119862 = 0 if 119868 is in the dashed rectangleminus3 if 119868 is out of this rectangle

(3)

minus3 is chosen experimentally according to simulationsSince the transmission distance of wireless LAN card is

limited geographic distance of each hop is the key factor thataffects the quality of communication If the distance stays

L

R

I1

I2

Figure 8 Different mark for different regions

unchanged then the route is really stable so that the sourceand destination can communicate steadily If the distance isincreasing it is possible that one node in this routewill exceedthe transmission distance soon If the distance is decreasingthen after some time the two nodes may cross over and startto depart It is better than increasing distance but not asgood as unchanged distance So we set two values for eachchanged distance 0 for each decreasing distance and minus2 foreach increasing distance as shown in (2)

To find a path to the destination the source broadcastsan RREQ containing the position and velocity of source andpreset a standard mark value

As illustrated in Figure 9 when a node receives an RREQthe following will occur

(i) If it does not have a route to the source node it willinsert the route RREQ into its routing table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination or has a route to destination itwill create an RREP in which the mark is a standardvalue (if it is the destination) or is obtained fromrouting table (if it has a route to destination)

(iv) If it is a node besides (3) and satisfies the broadcastingconditions (refer to flooding rules) it will add themark of sourcerarr precursorrarr current to the presentmark in RREQ It will also update the position andvelocity of current node in RREQ

As illustrated in Figure 10 when a node receives an RREPthe following will occur

(i) If it does not have a route to the destination it willinsert the route which RREP goes through into itsrouting table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination of RREP (ie the source ofRREQ) it will insert the route into routing table

(iv) If it is a forwarding node before unicasting RREQit will add the mark of sourcerarr precursorrarr current

International Journal of Distributed Sensor Networks 9

Receives anRREQ

Has a route tosource

Inserts inrouting table

NeedsupdateUpdates

Destination or has aroute to destination

Replies anRREP

Needsforward

DiscardsRREQ

Updates markand current

nodersquos positionand velocity in

RREQ

BroadcastsRREQ

No

Yes

Yes

No

YesNo

No

Yes

Figure 9 Flowchart of processing RREQ

to the present mark in RREP It will also update theposition and velocity of current node in RREQ

44 Simulation Setup

(1) Simulation Tools We chose VanetMobiSim 11 [30] fortraffic simulation This software can generate a traffic flow inthe format suitable forNS2 [31] which can loadGBAODV fornetwork simulation

With VanetMobiSim we import maps from the USCensus Bureau TIGERLine database [32] which includescomplete coverage of the United States Puerto Rico and soforthMoreoverVanetMobiSim supports formultilane roadsdifferentiated speed constraints and traffic light signals atintersections All the vehicles can be set to Intelligent DriverModel with Lane Changing (IDM LC) [33 34] For thesereasons the scenario in traffic layer is quite authentic whichmakes the simulation in network layer reliable

(2) Parameter Settings In our simulation we observed twotypes of traffic models downtown and highway TGR11001[32] (district of Columbia WA) is chosen as downtown

Receives anRREP

Has a route todestination

Inserts inrouting table

NeedsupdateUpdates

Inserts inrouting table

Sourcenode

Update mark andcurrent nodersquosposition and

velocity in RREP

Unicast RREP

NoYes

Yes

No

Yes

No

Figure 10 Flowchart of processing RREP

Table 4 Parameter settings of traffic

Traffic layer Downtown HighwayArea (m2) 1000 times 1000 1000 times 1000

Number of lanes 3 3Maximum number of traffic lights 10 NoneSpeed (kmh) 20sim80 60sim120Simulation time (s) 250 1000

Table 5 Parameter settings of network

Network layer Downtown HighwayMaximum transmission distance (m) 250 250Number of sources 35 17Number of connections 56 26CBR packet size (bytes) 256 512Transmission rate (pkts) 1 2Simulation time (s) 250 1000

map and TGR36001 [32] (Albany county NY) is chosen ashighway map Multilanes and traffic lights are involved Allthe vehicles follow the IDM LC driving model Tables 4 and5 are the parameter settings of traffic and network simulationTraffic flow and CBR (constant bitrate) data flow are bothgenerated randomly

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 5: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of Distributed Sensor Networks 5

0

001

002

003

004

005

006

007

Pack

et d

elay

(s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

0 1000 2000 3000 4000 5000 6000 70000

005

015

01

025

02

035

03

Packet ID

Pack

et d

elay

(s)

(b) DSR

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

008

Packet ID

Pack

et d

elay

(s)

(c) AODV

Figure 4 Packet delay of three routing protocols

if an acknowledgement is not received Consequently thisincreases packet delay since the packet must wait until a newroute is established(3) Packet Jitter Figure 5 shows the jitter of data packetsusing three different routing protocols The statistical valuesof them are (2356 plusmn 4199)ms (DSDV 120572 = 095) (3254 plusmn

10630)ms (DSR 120572 = 095) and (4498 plusmn 4498)ms (AODV120572 = 095) respectively DSDV shows the most superioritythan others The average jitter of AODV is the biggest amongthe three protocols Moreover the standard deviation of jitterin AODV is larger than that in DSDV Although the averagejitter of DSR is not the largest among the three protocolsDSR has the largest jitter fluctuation among themThismeansviolent variation exists in the delay of a few data packets sentin DSRThe rapid change of jitter is attributed to the frequentchange of network topology and the mechanism of inherentrouting update(4) Throughput Figure 6 shows the throughput compari-son of DSDV DSR and AODV All throughputs are ever-increasing with the time in general This can be attributed

to more active nodes to join network communication withtime extensionThe graph also reveals that AODV has higherthroughput than DSR In DSR a route is chosen basedon the short delay at the instance of route establishmentAlthough this path may be the best route at that instantit may be also a route that lacks routing stability or hasunacceptably high load In contrast AODVhas amore robustupdate mechanism to avoid bottleneck and congestion andeventually improve throughput In DSDV high packet lossrate causes throughput to drop

(5) Protocol Overhead Table 3 shows the overhead compar-ison of DSDV AODV and DSR AODV has the highestoverhead among three routing protocols due to three mainreasons Firstly AODV allows broadcast Although the dis-covery packets are broadcasted only when necessary suchas establishing a new route link breakage or route errorthe broadcast instances will often appear for a fast mobileVANET Secondly AODV allows mobile nodes to respondto link breakages and changes in network topology in a

6 International Journal of Distributed Sensor Networks

0

001

002

003

004

005

006

007

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

0

005

015

01

025

02

035

03

(b) DSR

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

Packet ID

(c) AODV

Figure 5 Packet jitter of three routing protocols

timely mannerThus numerous routingmessages are used tomaintain an active route in AODVThirdly AODV is derivedfrom DSDV It still has similar features of proactive routingprotocols When the network topology is often changedbecause of the fast mobility of nodes proactive protocolsmust send more messages to maintain a valid routing tableThe experimental results are shown in Table 3 DSR hasthe best protocol overhead performance among the threeprotocols The DSR protocol is composed of the two mainmechanisms of ldquoroute discoveryrdquo and ldquoroute maintenancerdquowhich work together to allow nodes to discover andmaintainroutes to arbitrary destinations in the ad hoc network Somemeasures reducing overhead are adopted in the process ofldquoRoute Discoveryrdquo and ldquoRoute Maintenancerdquo of DSR [21]

34 Conclusion Based on the simulation results we discov-ered that DSDV achieves marginal better packet delay andpacket jitter than AODV and DSR However it has signif-icantly higher packet loss rate and the smallest throughoutamong them Although DSR has the best protocol overhead

Table 3 Overhead of three routing protocols

Type DSDV AODV DSRNumber of received data packets 5579 5999 5993Number of sent message packets 25607 39488 1615Protocol overhead 459 658 027

performance among three routing protocols it has poorerjitter and throughput thanAODV As a whole AODV ismoreappropriate for the freeway VANET according to the qualityof service and the real time of packet delivery The inherentreason producing this result is the effect of node mobility onthe performance of the routing protocol

4 GPS Based AODV Protocol forV2V Communication [22]

As we can find through the previous protocol evaluationAODV is appropriate for the dynamic structure with mobilevehicles The source node will find a route to destination

International Journal of Distributed Sensor Networks 7

0 50 100 150 2000

20

40

60

80

100

120

140

Time (s)

Thro

ughp

ut (k

bps)

DSRDSDVAODV

Figure 6 Throughput comparison of three routing protocols

for data transmission To find a route to the destination thesource broadcasts a route request packet (RREQ) The nodereceiving RREQ will reply a route reply packet (RREP) to thereverse path that RREQ went through if it is the destinationof RREQ or if it has a recent route to the destinationOtherwise the node will broadcast RREQ until either ofthe two situations above occurs As RREP traverses backto the source the nodes along the path enter the forwardroute into their routing tables If a node leaves the networkthe node that discovers this broken link will send a linkfailure notification (RERR) to the precursorsThe RERR goesupstreamuntil it reaches the sourceThe sourcewill thereafterrestart route discovery process if needed [23]

However AODV is dedicated forMANET andmodifica-tion is necessary for VANET applications [24] Since nodes inVANET are fast moving vehicles the topology of the networkchanges all the time making the position and velocity ofthe node critical factors while finding the routes Thereforebesides conventional topology-based routing protocol (suchas AODV) position-based routing protocol which primarilyuses the position information obtained byGPS to find a routeis applied in VANET [25]

In recent years many researchers have tried to combinetwo types of routing together In [26] Kim et al pro-posed AODV-RRS which restricts the number of forward-ing RREQs according to the concepts of stable zone andcaution zone PAODV [27] restricts the number of floodingRREQs based on the distance between current node andits neighbors Both of them have similar mechanism andreduce the number of broken links DAODV [28] establishesa route depending on the direction and position of sourceintermediate and destination node Although it also reducesthe number of broken links DAODV assumes that sourcenode knows the direction and position of all the nodes inthe network Actually this is impossible with the current GPS

devices We should find some mechanisms to get the geo-graphic information of all the nodesThereby in [29] Asenovand Hnatyshin proposed GeoAODV routing protocol Eachnode will maintain an additional table geotable to keep trackof the geographic information of all the nodes RREQs arerestricted flooding in the region determined by the geotableHowever it costs more resources to maintain two tables

We go a step further in this section by proposing aGPS based AODV (GBAODV) which enhances the overallperformance of AODV in VANET In order to be attainablein physical implementation GBAODV assumes that eachnode only knows its own position and velocity Withoutmaintaining a geotable [29] GBAODV is muchmore concisethan GeoAODV

41 Overview of GBAODV With current GPS device wecan obtain the longitude (119909 coordinate) and latitude (119910coordinate) of current node and calculate the speed in eachdirection with two successive sets of coordinates Withoutmaintaining a geotable we add geographic information intorouting table to make the algorithm concise

There are two main features in GBAODV First is toreduce the number of RREQs The node receiving an RREQpacket will check the distance and motion trend betweenthe precursor and itself so as to decide if this RREQ shouldbe broadcasted By restricting these flooding RREQs it willavoid lots of packet collisions in the network As a resultthe packet delivery ratio and throughput of the networkcan be raised Second is to mark the route according to thepositions and velocities of source intermediate and desti-nation Higher mark means higher stability Consequentlyevery node will choose a route with higher mark

These two features require specifying flooding rules andmarking standards as well as some modifications in therouting table and routing packets

42 Modified Structure of Routing Table RREQ and RREP

(1) Modified Routing Table Original routing table containsthe IP addresses of next hop and destination sequencenumber of destination and total hops The routing table willupdate if the sequence number of destination is larger ortotal hops is less We added the position and velocity of thedestination node and the mark of this route into the routingtable Besides the two conditions above the routing table willupdate when the mark is larger which means that the routepath is more stable

(2) Modified Frame Structure of RREQ We inserted position(119909 119910 coordinates) and velocity (119909 119910 direction) of the sourceand current node into RREQ packet We also applied areserved segment to record the mark of the route link fromsource to precursor

(3) Modified Frame Structure of RREP RREP is modifiedsimilar to RREQ We inserted position and velocity of thedestination and current node into RREP packet We also

8 International Journal of Distributed Sensor Networks

P

B

A

C

DE

04 lowast R

08 lowast R

Figure 7 B C and D will broadcast RREQ received from P

applied a reserved segment to record the mark of the routefrom destination to current node

43 Flooding Rules and Marking Standards

(1) Flooding Rules Let 119889 stand for the distance between theprecursor and current node and 119877 stand for the maximumtransmission distance of the wireless cardThe flooding rulesare as follows

If (119889 lt 04 119877 (119889 gt 08119877ampamp119889 is increasing))

discard RREQ

Else

broadcast RREQ

The coefficients 04 and 08 are chosen empiricallyWe conducted simulations with the combinations of thesecoefficients to study numbers of RREQs and RERRs packetloss ratio and average end-to-end delay Results show that thecombination of 04 and 08 is the best among all

In Figure 7 B C andDwill broadcast the RREQ receivedfrom P while A and E will discard it

(2) Marking Standards As illustrated in Figure 8 let 119871119863119888

stand for the current distance between node 119871 and node 119868119871119863119897stand for this distance a moment later 119877119863

119888stand for the

current distance between node 119868 and node 119877 and 119877119863119897stand

for this distance a moment later We specify the markingstandards as follows

mark = 2 sgn (119871119863119888minus 119871119863119897) + 2 sgn (119877119863

119888minus 119877119863119897) + 119862 (2)

where sgn() is a signum function119862 is a constant

119862 = 0 if 119868 is in the dashed rectangleminus3 if 119868 is out of this rectangle

(3)

minus3 is chosen experimentally according to simulationsSince the transmission distance of wireless LAN card is

limited geographic distance of each hop is the key factor thataffects the quality of communication If the distance stays

L

R

I1

I2

Figure 8 Different mark for different regions

unchanged then the route is really stable so that the sourceand destination can communicate steadily If the distance isincreasing it is possible that one node in this routewill exceedthe transmission distance soon If the distance is decreasingthen after some time the two nodes may cross over and startto depart It is better than increasing distance but not asgood as unchanged distance So we set two values for eachchanged distance 0 for each decreasing distance and minus2 foreach increasing distance as shown in (2)

To find a path to the destination the source broadcastsan RREQ containing the position and velocity of source andpreset a standard mark value

As illustrated in Figure 9 when a node receives an RREQthe following will occur

(i) If it does not have a route to the source node it willinsert the route RREQ into its routing table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination or has a route to destination itwill create an RREP in which the mark is a standardvalue (if it is the destination) or is obtained fromrouting table (if it has a route to destination)

(iv) If it is a node besides (3) and satisfies the broadcastingconditions (refer to flooding rules) it will add themark of sourcerarr precursorrarr current to the presentmark in RREQ It will also update the position andvelocity of current node in RREQ

As illustrated in Figure 10 when a node receives an RREPthe following will occur

(i) If it does not have a route to the destination it willinsert the route which RREP goes through into itsrouting table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination of RREP (ie the source ofRREQ) it will insert the route into routing table

(iv) If it is a forwarding node before unicasting RREQit will add the mark of sourcerarr precursorrarr current

International Journal of Distributed Sensor Networks 9

Receives anRREQ

Has a route tosource

Inserts inrouting table

NeedsupdateUpdates

Destination or has aroute to destination

Replies anRREP

Needsforward

DiscardsRREQ

Updates markand current

nodersquos positionand velocity in

RREQ

BroadcastsRREQ

No

Yes

Yes

No

YesNo

No

Yes

Figure 9 Flowchart of processing RREQ

to the present mark in RREP It will also update theposition and velocity of current node in RREQ

44 Simulation Setup

(1) Simulation Tools We chose VanetMobiSim 11 [30] fortraffic simulation This software can generate a traffic flow inthe format suitable forNS2 [31] which can loadGBAODV fornetwork simulation

With VanetMobiSim we import maps from the USCensus Bureau TIGERLine database [32] which includescomplete coverage of the United States Puerto Rico and soforthMoreoverVanetMobiSim supports formultilane roadsdifferentiated speed constraints and traffic light signals atintersections All the vehicles can be set to Intelligent DriverModel with Lane Changing (IDM LC) [33 34] For thesereasons the scenario in traffic layer is quite authentic whichmakes the simulation in network layer reliable

(2) Parameter Settings In our simulation we observed twotypes of traffic models downtown and highway TGR11001[32] (district of Columbia WA) is chosen as downtown

Receives anRREP

Has a route todestination

Inserts inrouting table

NeedsupdateUpdates

Inserts inrouting table

Sourcenode

Update mark andcurrent nodersquosposition and

velocity in RREP

Unicast RREP

NoYes

Yes

No

Yes

No

Figure 10 Flowchart of processing RREP

Table 4 Parameter settings of traffic

Traffic layer Downtown HighwayArea (m2) 1000 times 1000 1000 times 1000

Number of lanes 3 3Maximum number of traffic lights 10 NoneSpeed (kmh) 20sim80 60sim120Simulation time (s) 250 1000

Table 5 Parameter settings of network

Network layer Downtown HighwayMaximum transmission distance (m) 250 250Number of sources 35 17Number of connections 56 26CBR packet size (bytes) 256 512Transmission rate (pkts) 1 2Simulation time (s) 250 1000

map and TGR36001 [32] (Albany county NY) is chosen ashighway map Multilanes and traffic lights are involved Allthe vehicles follow the IDM LC driving model Tables 4 and5 are the parameter settings of traffic and network simulationTraffic flow and CBR (constant bitrate) data flow are bothgenerated randomly

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 6: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

6 International Journal of Distributed Sensor Networks

0

001

002

003

004

005

006

007

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

(a) DSDV

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 6000 7000Packet ID

0

005

015

01

025

02

035

03

(b) DSR

Pack

et ji

tter (

s)

0 1000 2000 3000 4000 5000 60000

001

002

003

004

005

006

007

Packet ID

(c) AODV

Figure 5 Packet jitter of three routing protocols

timely mannerThus numerous routingmessages are used tomaintain an active route in AODVThirdly AODV is derivedfrom DSDV It still has similar features of proactive routingprotocols When the network topology is often changedbecause of the fast mobility of nodes proactive protocolsmust send more messages to maintain a valid routing tableThe experimental results are shown in Table 3 DSR hasthe best protocol overhead performance among the threeprotocols The DSR protocol is composed of the two mainmechanisms of ldquoroute discoveryrdquo and ldquoroute maintenancerdquowhich work together to allow nodes to discover andmaintainroutes to arbitrary destinations in the ad hoc network Somemeasures reducing overhead are adopted in the process ofldquoRoute Discoveryrdquo and ldquoRoute Maintenancerdquo of DSR [21]

34 Conclusion Based on the simulation results we discov-ered that DSDV achieves marginal better packet delay andpacket jitter than AODV and DSR However it has signif-icantly higher packet loss rate and the smallest throughoutamong them Although DSR has the best protocol overhead

Table 3 Overhead of three routing protocols

Type DSDV AODV DSRNumber of received data packets 5579 5999 5993Number of sent message packets 25607 39488 1615Protocol overhead 459 658 027

performance among three routing protocols it has poorerjitter and throughput thanAODV As a whole AODV ismoreappropriate for the freeway VANET according to the qualityof service and the real time of packet delivery The inherentreason producing this result is the effect of node mobility onthe performance of the routing protocol

4 GPS Based AODV Protocol forV2V Communication [22]

As we can find through the previous protocol evaluationAODV is appropriate for the dynamic structure with mobilevehicles The source node will find a route to destination

International Journal of Distributed Sensor Networks 7

0 50 100 150 2000

20

40

60

80

100

120

140

Time (s)

Thro

ughp

ut (k

bps)

DSRDSDVAODV

Figure 6 Throughput comparison of three routing protocols

for data transmission To find a route to the destination thesource broadcasts a route request packet (RREQ) The nodereceiving RREQ will reply a route reply packet (RREP) to thereverse path that RREQ went through if it is the destinationof RREQ or if it has a recent route to the destinationOtherwise the node will broadcast RREQ until either ofthe two situations above occurs As RREP traverses backto the source the nodes along the path enter the forwardroute into their routing tables If a node leaves the networkthe node that discovers this broken link will send a linkfailure notification (RERR) to the precursorsThe RERR goesupstreamuntil it reaches the sourceThe sourcewill thereafterrestart route discovery process if needed [23]

However AODV is dedicated forMANET andmodifica-tion is necessary for VANET applications [24] Since nodes inVANET are fast moving vehicles the topology of the networkchanges all the time making the position and velocity ofthe node critical factors while finding the routes Thereforebesides conventional topology-based routing protocol (suchas AODV) position-based routing protocol which primarilyuses the position information obtained byGPS to find a routeis applied in VANET [25]

In recent years many researchers have tried to combinetwo types of routing together In [26] Kim et al pro-posed AODV-RRS which restricts the number of forward-ing RREQs according to the concepts of stable zone andcaution zone PAODV [27] restricts the number of floodingRREQs based on the distance between current node andits neighbors Both of them have similar mechanism andreduce the number of broken links DAODV [28] establishesa route depending on the direction and position of sourceintermediate and destination node Although it also reducesthe number of broken links DAODV assumes that sourcenode knows the direction and position of all the nodes inthe network Actually this is impossible with the current GPS

devices We should find some mechanisms to get the geo-graphic information of all the nodesThereby in [29] Asenovand Hnatyshin proposed GeoAODV routing protocol Eachnode will maintain an additional table geotable to keep trackof the geographic information of all the nodes RREQs arerestricted flooding in the region determined by the geotableHowever it costs more resources to maintain two tables

We go a step further in this section by proposing aGPS based AODV (GBAODV) which enhances the overallperformance of AODV in VANET In order to be attainablein physical implementation GBAODV assumes that eachnode only knows its own position and velocity Withoutmaintaining a geotable [29] GBAODV is muchmore concisethan GeoAODV

41 Overview of GBAODV With current GPS device wecan obtain the longitude (119909 coordinate) and latitude (119910coordinate) of current node and calculate the speed in eachdirection with two successive sets of coordinates Withoutmaintaining a geotable we add geographic information intorouting table to make the algorithm concise

There are two main features in GBAODV First is toreduce the number of RREQs The node receiving an RREQpacket will check the distance and motion trend betweenthe precursor and itself so as to decide if this RREQ shouldbe broadcasted By restricting these flooding RREQs it willavoid lots of packet collisions in the network As a resultthe packet delivery ratio and throughput of the networkcan be raised Second is to mark the route according to thepositions and velocities of source intermediate and desti-nation Higher mark means higher stability Consequentlyevery node will choose a route with higher mark

These two features require specifying flooding rules andmarking standards as well as some modifications in therouting table and routing packets

42 Modified Structure of Routing Table RREQ and RREP

(1) Modified Routing Table Original routing table containsthe IP addresses of next hop and destination sequencenumber of destination and total hops The routing table willupdate if the sequence number of destination is larger ortotal hops is less We added the position and velocity of thedestination node and the mark of this route into the routingtable Besides the two conditions above the routing table willupdate when the mark is larger which means that the routepath is more stable

(2) Modified Frame Structure of RREQ We inserted position(119909 119910 coordinates) and velocity (119909 119910 direction) of the sourceand current node into RREQ packet We also applied areserved segment to record the mark of the route link fromsource to precursor

(3) Modified Frame Structure of RREP RREP is modifiedsimilar to RREQ We inserted position and velocity of thedestination and current node into RREP packet We also

8 International Journal of Distributed Sensor Networks

P

B

A

C

DE

04 lowast R

08 lowast R

Figure 7 B C and D will broadcast RREQ received from P

applied a reserved segment to record the mark of the routefrom destination to current node

43 Flooding Rules and Marking Standards

(1) Flooding Rules Let 119889 stand for the distance between theprecursor and current node and 119877 stand for the maximumtransmission distance of the wireless cardThe flooding rulesare as follows

If (119889 lt 04 119877 (119889 gt 08119877ampamp119889 is increasing))

discard RREQ

Else

broadcast RREQ

The coefficients 04 and 08 are chosen empiricallyWe conducted simulations with the combinations of thesecoefficients to study numbers of RREQs and RERRs packetloss ratio and average end-to-end delay Results show that thecombination of 04 and 08 is the best among all

In Figure 7 B C andDwill broadcast the RREQ receivedfrom P while A and E will discard it

(2) Marking Standards As illustrated in Figure 8 let 119871119863119888

stand for the current distance between node 119871 and node 119868119871119863119897stand for this distance a moment later 119877119863

119888stand for the

current distance between node 119868 and node 119877 and 119877119863119897stand

for this distance a moment later We specify the markingstandards as follows

mark = 2 sgn (119871119863119888minus 119871119863119897) + 2 sgn (119877119863

119888minus 119877119863119897) + 119862 (2)

where sgn() is a signum function119862 is a constant

119862 = 0 if 119868 is in the dashed rectangleminus3 if 119868 is out of this rectangle

(3)

minus3 is chosen experimentally according to simulationsSince the transmission distance of wireless LAN card is

limited geographic distance of each hop is the key factor thataffects the quality of communication If the distance stays

L

R

I1

I2

Figure 8 Different mark for different regions

unchanged then the route is really stable so that the sourceand destination can communicate steadily If the distance isincreasing it is possible that one node in this routewill exceedthe transmission distance soon If the distance is decreasingthen after some time the two nodes may cross over and startto depart It is better than increasing distance but not asgood as unchanged distance So we set two values for eachchanged distance 0 for each decreasing distance and minus2 foreach increasing distance as shown in (2)

To find a path to the destination the source broadcastsan RREQ containing the position and velocity of source andpreset a standard mark value

As illustrated in Figure 9 when a node receives an RREQthe following will occur

(i) If it does not have a route to the source node it willinsert the route RREQ into its routing table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination or has a route to destination itwill create an RREP in which the mark is a standardvalue (if it is the destination) or is obtained fromrouting table (if it has a route to destination)

(iv) If it is a node besides (3) and satisfies the broadcastingconditions (refer to flooding rules) it will add themark of sourcerarr precursorrarr current to the presentmark in RREQ It will also update the position andvelocity of current node in RREQ

As illustrated in Figure 10 when a node receives an RREPthe following will occur

(i) If it does not have a route to the destination it willinsert the route which RREP goes through into itsrouting table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination of RREP (ie the source ofRREQ) it will insert the route into routing table

(iv) If it is a forwarding node before unicasting RREQit will add the mark of sourcerarr precursorrarr current

International Journal of Distributed Sensor Networks 9

Receives anRREQ

Has a route tosource

Inserts inrouting table

NeedsupdateUpdates

Destination or has aroute to destination

Replies anRREP

Needsforward

DiscardsRREQ

Updates markand current

nodersquos positionand velocity in

RREQ

BroadcastsRREQ

No

Yes

Yes

No

YesNo

No

Yes

Figure 9 Flowchart of processing RREQ

to the present mark in RREP It will also update theposition and velocity of current node in RREQ

44 Simulation Setup

(1) Simulation Tools We chose VanetMobiSim 11 [30] fortraffic simulation This software can generate a traffic flow inthe format suitable forNS2 [31] which can loadGBAODV fornetwork simulation

With VanetMobiSim we import maps from the USCensus Bureau TIGERLine database [32] which includescomplete coverage of the United States Puerto Rico and soforthMoreoverVanetMobiSim supports formultilane roadsdifferentiated speed constraints and traffic light signals atintersections All the vehicles can be set to Intelligent DriverModel with Lane Changing (IDM LC) [33 34] For thesereasons the scenario in traffic layer is quite authentic whichmakes the simulation in network layer reliable

(2) Parameter Settings In our simulation we observed twotypes of traffic models downtown and highway TGR11001[32] (district of Columbia WA) is chosen as downtown

Receives anRREP

Has a route todestination

Inserts inrouting table

NeedsupdateUpdates

Inserts inrouting table

Sourcenode

Update mark andcurrent nodersquosposition and

velocity in RREP

Unicast RREP

NoYes

Yes

No

Yes

No

Figure 10 Flowchart of processing RREP

Table 4 Parameter settings of traffic

Traffic layer Downtown HighwayArea (m2) 1000 times 1000 1000 times 1000

Number of lanes 3 3Maximum number of traffic lights 10 NoneSpeed (kmh) 20sim80 60sim120Simulation time (s) 250 1000

Table 5 Parameter settings of network

Network layer Downtown HighwayMaximum transmission distance (m) 250 250Number of sources 35 17Number of connections 56 26CBR packet size (bytes) 256 512Transmission rate (pkts) 1 2Simulation time (s) 250 1000

map and TGR36001 [32] (Albany county NY) is chosen ashighway map Multilanes and traffic lights are involved Allthe vehicles follow the IDM LC driving model Tables 4 and5 are the parameter settings of traffic and network simulationTraffic flow and CBR (constant bitrate) data flow are bothgenerated randomly

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 7: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of Distributed Sensor Networks 7

0 50 100 150 2000

20

40

60

80

100

120

140

Time (s)

Thro

ughp

ut (k

bps)

DSRDSDVAODV

Figure 6 Throughput comparison of three routing protocols

for data transmission To find a route to the destination thesource broadcasts a route request packet (RREQ) The nodereceiving RREQ will reply a route reply packet (RREP) to thereverse path that RREQ went through if it is the destinationof RREQ or if it has a recent route to the destinationOtherwise the node will broadcast RREQ until either ofthe two situations above occurs As RREP traverses backto the source the nodes along the path enter the forwardroute into their routing tables If a node leaves the networkthe node that discovers this broken link will send a linkfailure notification (RERR) to the precursorsThe RERR goesupstreamuntil it reaches the sourceThe sourcewill thereafterrestart route discovery process if needed [23]

However AODV is dedicated forMANET andmodifica-tion is necessary for VANET applications [24] Since nodes inVANET are fast moving vehicles the topology of the networkchanges all the time making the position and velocity ofthe node critical factors while finding the routes Thereforebesides conventional topology-based routing protocol (suchas AODV) position-based routing protocol which primarilyuses the position information obtained byGPS to find a routeis applied in VANET [25]

In recent years many researchers have tried to combinetwo types of routing together In [26] Kim et al pro-posed AODV-RRS which restricts the number of forward-ing RREQs according to the concepts of stable zone andcaution zone PAODV [27] restricts the number of floodingRREQs based on the distance between current node andits neighbors Both of them have similar mechanism andreduce the number of broken links DAODV [28] establishesa route depending on the direction and position of sourceintermediate and destination node Although it also reducesthe number of broken links DAODV assumes that sourcenode knows the direction and position of all the nodes inthe network Actually this is impossible with the current GPS

devices We should find some mechanisms to get the geo-graphic information of all the nodesThereby in [29] Asenovand Hnatyshin proposed GeoAODV routing protocol Eachnode will maintain an additional table geotable to keep trackof the geographic information of all the nodes RREQs arerestricted flooding in the region determined by the geotableHowever it costs more resources to maintain two tables

We go a step further in this section by proposing aGPS based AODV (GBAODV) which enhances the overallperformance of AODV in VANET In order to be attainablein physical implementation GBAODV assumes that eachnode only knows its own position and velocity Withoutmaintaining a geotable [29] GBAODV is muchmore concisethan GeoAODV

41 Overview of GBAODV With current GPS device wecan obtain the longitude (119909 coordinate) and latitude (119910coordinate) of current node and calculate the speed in eachdirection with two successive sets of coordinates Withoutmaintaining a geotable we add geographic information intorouting table to make the algorithm concise

There are two main features in GBAODV First is toreduce the number of RREQs The node receiving an RREQpacket will check the distance and motion trend betweenthe precursor and itself so as to decide if this RREQ shouldbe broadcasted By restricting these flooding RREQs it willavoid lots of packet collisions in the network As a resultthe packet delivery ratio and throughput of the networkcan be raised Second is to mark the route according to thepositions and velocities of source intermediate and desti-nation Higher mark means higher stability Consequentlyevery node will choose a route with higher mark

These two features require specifying flooding rules andmarking standards as well as some modifications in therouting table and routing packets

42 Modified Structure of Routing Table RREQ and RREP

(1) Modified Routing Table Original routing table containsthe IP addresses of next hop and destination sequencenumber of destination and total hops The routing table willupdate if the sequence number of destination is larger ortotal hops is less We added the position and velocity of thedestination node and the mark of this route into the routingtable Besides the two conditions above the routing table willupdate when the mark is larger which means that the routepath is more stable

(2) Modified Frame Structure of RREQ We inserted position(119909 119910 coordinates) and velocity (119909 119910 direction) of the sourceand current node into RREQ packet We also applied areserved segment to record the mark of the route link fromsource to precursor

(3) Modified Frame Structure of RREP RREP is modifiedsimilar to RREQ We inserted position and velocity of thedestination and current node into RREP packet We also

8 International Journal of Distributed Sensor Networks

P

B

A

C

DE

04 lowast R

08 lowast R

Figure 7 B C and D will broadcast RREQ received from P

applied a reserved segment to record the mark of the routefrom destination to current node

43 Flooding Rules and Marking Standards

(1) Flooding Rules Let 119889 stand for the distance between theprecursor and current node and 119877 stand for the maximumtransmission distance of the wireless cardThe flooding rulesare as follows

If (119889 lt 04 119877 (119889 gt 08119877ampamp119889 is increasing))

discard RREQ

Else

broadcast RREQ

The coefficients 04 and 08 are chosen empiricallyWe conducted simulations with the combinations of thesecoefficients to study numbers of RREQs and RERRs packetloss ratio and average end-to-end delay Results show that thecombination of 04 and 08 is the best among all

In Figure 7 B C andDwill broadcast the RREQ receivedfrom P while A and E will discard it

(2) Marking Standards As illustrated in Figure 8 let 119871119863119888

stand for the current distance between node 119871 and node 119868119871119863119897stand for this distance a moment later 119877119863

119888stand for the

current distance between node 119868 and node 119877 and 119877119863119897stand

for this distance a moment later We specify the markingstandards as follows

mark = 2 sgn (119871119863119888minus 119871119863119897) + 2 sgn (119877119863

119888minus 119877119863119897) + 119862 (2)

where sgn() is a signum function119862 is a constant

119862 = 0 if 119868 is in the dashed rectangleminus3 if 119868 is out of this rectangle

(3)

minus3 is chosen experimentally according to simulationsSince the transmission distance of wireless LAN card is

limited geographic distance of each hop is the key factor thataffects the quality of communication If the distance stays

L

R

I1

I2

Figure 8 Different mark for different regions

unchanged then the route is really stable so that the sourceand destination can communicate steadily If the distance isincreasing it is possible that one node in this routewill exceedthe transmission distance soon If the distance is decreasingthen after some time the two nodes may cross over and startto depart It is better than increasing distance but not asgood as unchanged distance So we set two values for eachchanged distance 0 for each decreasing distance and minus2 foreach increasing distance as shown in (2)

To find a path to the destination the source broadcastsan RREQ containing the position and velocity of source andpreset a standard mark value

As illustrated in Figure 9 when a node receives an RREQthe following will occur

(i) If it does not have a route to the source node it willinsert the route RREQ into its routing table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination or has a route to destination itwill create an RREP in which the mark is a standardvalue (if it is the destination) or is obtained fromrouting table (if it has a route to destination)

(iv) If it is a node besides (3) and satisfies the broadcastingconditions (refer to flooding rules) it will add themark of sourcerarr precursorrarr current to the presentmark in RREQ It will also update the position andvelocity of current node in RREQ

As illustrated in Figure 10 when a node receives an RREPthe following will occur

(i) If it does not have a route to the destination it willinsert the route which RREP goes through into itsrouting table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination of RREP (ie the source ofRREQ) it will insert the route into routing table

(iv) If it is a forwarding node before unicasting RREQit will add the mark of sourcerarr precursorrarr current

International Journal of Distributed Sensor Networks 9

Receives anRREQ

Has a route tosource

Inserts inrouting table

NeedsupdateUpdates

Destination or has aroute to destination

Replies anRREP

Needsforward

DiscardsRREQ

Updates markand current

nodersquos positionand velocity in

RREQ

BroadcastsRREQ

No

Yes

Yes

No

YesNo

No

Yes

Figure 9 Flowchart of processing RREQ

to the present mark in RREP It will also update theposition and velocity of current node in RREQ

44 Simulation Setup

(1) Simulation Tools We chose VanetMobiSim 11 [30] fortraffic simulation This software can generate a traffic flow inthe format suitable forNS2 [31] which can loadGBAODV fornetwork simulation

With VanetMobiSim we import maps from the USCensus Bureau TIGERLine database [32] which includescomplete coverage of the United States Puerto Rico and soforthMoreoverVanetMobiSim supports formultilane roadsdifferentiated speed constraints and traffic light signals atintersections All the vehicles can be set to Intelligent DriverModel with Lane Changing (IDM LC) [33 34] For thesereasons the scenario in traffic layer is quite authentic whichmakes the simulation in network layer reliable

(2) Parameter Settings In our simulation we observed twotypes of traffic models downtown and highway TGR11001[32] (district of Columbia WA) is chosen as downtown

Receives anRREP

Has a route todestination

Inserts inrouting table

NeedsupdateUpdates

Inserts inrouting table

Sourcenode

Update mark andcurrent nodersquosposition and

velocity in RREP

Unicast RREP

NoYes

Yes

No

Yes

No

Figure 10 Flowchart of processing RREP

Table 4 Parameter settings of traffic

Traffic layer Downtown HighwayArea (m2) 1000 times 1000 1000 times 1000

Number of lanes 3 3Maximum number of traffic lights 10 NoneSpeed (kmh) 20sim80 60sim120Simulation time (s) 250 1000

Table 5 Parameter settings of network

Network layer Downtown HighwayMaximum transmission distance (m) 250 250Number of sources 35 17Number of connections 56 26CBR packet size (bytes) 256 512Transmission rate (pkts) 1 2Simulation time (s) 250 1000

map and TGR36001 [32] (Albany county NY) is chosen ashighway map Multilanes and traffic lights are involved Allthe vehicles follow the IDM LC driving model Tables 4 and5 are the parameter settings of traffic and network simulationTraffic flow and CBR (constant bitrate) data flow are bothgenerated randomly

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 8: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

8 International Journal of Distributed Sensor Networks

P

B

A

C

DE

04 lowast R

08 lowast R

Figure 7 B C and D will broadcast RREQ received from P

applied a reserved segment to record the mark of the routefrom destination to current node

43 Flooding Rules and Marking Standards

(1) Flooding Rules Let 119889 stand for the distance between theprecursor and current node and 119877 stand for the maximumtransmission distance of the wireless cardThe flooding rulesare as follows

If (119889 lt 04 119877 (119889 gt 08119877ampamp119889 is increasing))

discard RREQ

Else

broadcast RREQ

The coefficients 04 and 08 are chosen empiricallyWe conducted simulations with the combinations of thesecoefficients to study numbers of RREQs and RERRs packetloss ratio and average end-to-end delay Results show that thecombination of 04 and 08 is the best among all

In Figure 7 B C andDwill broadcast the RREQ receivedfrom P while A and E will discard it

(2) Marking Standards As illustrated in Figure 8 let 119871119863119888

stand for the current distance between node 119871 and node 119868119871119863119897stand for this distance a moment later 119877119863

119888stand for the

current distance between node 119868 and node 119877 and 119877119863119897stand

for this distance a moment later We specify the markingstandards as follows

mark = 2 sgn (119871119863119888minus 119871119863119897) + 2 sgn (119877119863

119888minus 119877119863119897) + 119862 (2)

where sgn() is a signum function119862 is a constant

119862 = 0 if 119868 is in the dashed rectangleminus3 if 119868 is out of this rectangle

(3)

minus3 is chosen experimentally according to simulationsSince the transmission distance of wireless LAN card is

limited geographic distance of each hop is the key factor thataffects the quality of communication If the distance stays

L

R

I1

I2

Figure 8 Different mark for different regions

unchanged then the route is really stable so that the sourceand destination can communicate steadily If the distance isincreasing it is possible that one node in this routewill exceedthe transmission distance soon If the distance is decreasingthen after some time the two nodes may cross over and startto depart It is better than increasing distance but not asgood as unchanged distance So we set two values for eachchanged distance 0 for each decreasing distance and minus2 foreach increasing distance as shown in (2)

To find a path to the destination the source broadcastsan RREQ containing the position and velocity of source andpreset a standard mark value

As illustrated in Figure 9 when a node receives an RREQthe following will occur

(i) If it does not have a route to the source node it willinsert the route RREQ into its routing table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination or has a route to destination itwill create an RREP in which the mark is a standardvalue (if it is the destination) or is obtained fromrouting table (if it has a route to destination)

(iv) If it is a node besides (3) and satisfies the broadcastingconditions (refer to flooding rules) it will add themark of sourcerarr precursorrarr current to the presentmark in RREQ It will also update the position andvelocity of current node in RREQ

As illustrated in Figure 10 when a node receives an RREPthe following will occur

(i) If it does not have a route to the destination it willinsert the route which RREP goes through into itsrouting table

(ii) If it has a route in routing table but the route needs tobe updated it will update routing table

(iii) If it is the destination of RREP (ie the source ofRREQ) it will insert the route into routing table

(iv) If it is a forwarding node before unicasting RREQit will add the mark of sourcerarr precursorrarr current

International Journal of Distributed Sensor Networks 9

Receives anRREQ

Has a route tosource

Inserts inrouting table

NeedsupdateUpdates

Destination or has aroute to destination

Replies anRREP

Needsforward

DiscardsRREQ

Updates markand current

nodersquos positionand velocity in

RREQ

BroadcastsRREQ

No

Yes

Yes

No

YesNo

No

Yes

Figure 9 Flowchart of processing RREQ

to the present mark in RREP It will also update theposition and velocity of current node in RREQ

44 Simulation Setup

(1) Simulation Tools We chose VanetMobiSim 11 [30] fortraffic simulation This software can generate a traffic flow inthe format suitable forNS2 [31] which can loadGBAODV fornetwork simulation

With VanetMobiSim we import maps from the USCensus Bureau TIGERLine database [32] which includescomplete coverage of the United States Puerto Rico and soforthMoreoverVanetMobiSim supports formultilane roadsdifferentiated speed constraints and traffic light signals atintersections All the vehicles can be set to Intelligent DriverModel with Lane Changing (IDM LC) [33 34] For thesereasons the scenario in traffic layer is quite authentic whichmakes the simulation in network layer reliable

(2) Parameter Settings In our simulation we observed twotypes of traffic models downtown and highway TGR11001[32] (district of Columbia WA) is chosen as downtown

Receives anRREP

Has a route todestination

Inserts inrouting table

NeedsupdateUpdates

Inserts inrouting table

Sourcenode

Update mark andcurrent nodersquosposition and

velocity in RREP

Unicast RREP

NoYes

Yes

No

Yes

No

Figure 10 Flowchart of processing RREP

Table 4 Parameter settings of traffic

Traffic layer Downtown HighwayArea (m2) 1000 times 1000 1000 times 1000

Number of lanes 3 3Maximum number of traffic lights 10 NoneSpeed (kmh) 20sim80 60sim120Simulation time (s) 250 1000

Table 5 Parameter settings of network

Network layer Downtown HighwayMaximum transmission distance (m) 250 250Number of sources 35 17Number of connections 56 26CBR packet size (bytes) 256 512Transmission rate (pkts) 1 2Simulation time (s) 250 1000

map and TGR36001 [32] (Albany county NY) is chosen ashighway map Multilanes and traffic lights are involved Allthe vehicles follow the IDM LC driving model Tables 4 and5 are the parameter settings of traffic and network simulationTraffic flow and CBR (constant bitrate) data flow are bothgenerated randomly

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 9: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of Distributed Sensor Networks 9

Receives anRREQ

Has a route tosource

Inserts inrouting table

NeedsupdateUpdates

Destination or has aroute to destination

Replies anRREP

Needsforward

DiscardsRREQ

Updates markand current

nodersquos positionand velocity in

RREQ

BroadcastsRREQ

No

Yes

Yes

No

YesNo

No

Yes

Figure 9 Flowchart of processing RREQ

to the present mark in RREP It will also update theposition and velocity of current node in RREQ

44 Simulation Setup

(1) Simulation Tools We chose VanetMobiSim 11 [30] fortraffic simulation This software can generate a traffic flow inthe format suitable forNS2 [31] which can loadGBAODV fornetwork simulation

With VanetMobiSim we import maps from the USCensus Bureau TIGERLine database [32] which includescomplete coverage of the United States Puerto Rico and soforthMoreoverVanetMobiSim supports formultilane roadsdifferentiated speed constraints and traffic light signals atintersections All the vehicles can be set to Intelligent DriverModel with Lane Changing (IDM LC) [33 34] For thesereasons the scenario in traffic layer is quite authentic whichmakes the simulation in network layer reliable

(2) Parameter Settings In our simulation we observed twotypes of traffic models downtown and highway TGR11001[32] (district of Columbia WA) is chosen as downtown

Receives anRREP

Has a route todestination

Inserts inrouting table

NeedsupdateUpdates

Inserts inrouting table

Sourcenode

Update mark andcurrent nodersquosposition and

velocity in RREP

Unicast RREP

NoYes

Yes

No

Yes

No

Figure 10 Flowchart of processing RREP

Table 4 Parameter settings of traffic

Traffic layer Downtown HighwayArea (m2) 1000 times 1000 1000 times 1000

Number of lanes 3 3Maximum number of traffic lights 10 NoneSpeed (kmh) 20sim80 60sim120Simulation time (s) 250 1000

Table 5 Parameter settings of network

Network layer Downtown HighwayMaximum transmission distance (m) 250 250Number of sources 35 17Number of connections 56 26CBR packet size (bytes) 256 512Transmission rate (pkts) 1 2Simulation time (s) 250 1000

map and TGR36001 [32] (Albany county NY) is chosen ashighway map Multilanes and traffic lights are involved Allthe vehicles follow the IDM LC driving model Tables 4 and5 are the parameter settings of traffic and network simulationTraffic flow and CBR (constant bitrate) data flow are bothgenerated randomly

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 10: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

10 International Journal of Distributed Sensor Networks

01020304050607080

100 110 120 130 140

Num

ber o

f RRE

Qs

Number of nodes

Downtown

AODVGBAODV

Figure 11 Number of RREQs received downtown per node persecond

0002004006008

01012014

100 110 120 130 140

Num

ber o

f RER

Rs

Number of nodes

Downtown

AODVGBAODV

Figure 12 Number of RERRs sent in downtown scenario perconnection per second

45 Simulation Results

(1) Downtown Model Figure 11 illustrates that the number ofRREQs received per node per second is reduced by about50 This is caused by the application of flooding rules Inaddition we can notice that although it is normalized by thenumber of nodes the number of RREQs still increases withthe number of nodes This means larger number of nodesinduces larger amount of RREQs broadcasted in the wholenetworkTherefore it is significant to reduce the broadcastedRREQs especially in high density traffic

Figure 12 illustrates the number of RERRs sent per con-nection The number of RERRs is also reduced a lot whichmeans broken links have decreased a lot This is an attributeto the application ofmarking standards sincewe choose everyconnection with high stability

Figures 13 and 14 illustrate the packet loss ratio andaverage end-to-end delay Compared with Figures 11 and12 they show that packet loss ratio and average end-to-enddelay are positive correlated to the numbers of RREQs andRERRs because reducing the number of RREQs contributesto avoiding large amount of packet collisions in the network

0

01

02

03

04

100 110 120 130 140

Pack

ets l

oss r

atio

Number of nodes

Downtown

AODVGBAODV

Figure 13 Packet loss ratio in downtown scenario

0

100

200

300

400

500

600

100 110 120 130 140Number of nodes

Downtown

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 14 Average end-to-end delay in downtown scenario

Meanwhile reducing the number of RERRs (ie brokenlinks) could smooth communication

(2) Highway Model The number of RREQs received pernode per second is reduced by more than 50 (illustratedin Figure 15) Figure 16 shows that the number of RERRssent per connection is also reduced We also note that withthe increase of number of nodes the reduction of RERRs(ie broken links) increases This means GBAODV is moreefficient in high density traffic scenario The conclusion isverified by Figure 17 (packet loss ratio) and Figure 18 (averageend-to-end delay) Compared with Figures 13 and 14 theimprovement of network performance is not as sharp as thatwhich we obtained in downtown model

The main reason is that the node density in highwaymodel is relatively low Firstly lower node density leads toless number of RREQs flooding in the network (referringto Figures 11 and 15) The number of RREQs received pernode per second in highway model is about 80 of thatin downtown model Therefore packet collision in highwayis slighter than in downtown Although GBAODV weakenspacket collision it achieves no big improvement Secondlylower node density provides fewer choices of stable routesIf we restrict the number of RREQs the effect caused byskipping some stable routes is larger than that in downtown

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 11: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of Distributed Sensor Networks 11

010203040506070

50 60 70 80 90

Num

ber o

f RRE

Qs

Number of nodes

Highway

AODVGBAODV

Figure 15 Number of RREQs sent in highway scenario per node persecond

0002004006008

01012014

50 60 70 80 90

Num

ber o

f RER

Rs

Number of nodes

Highway

AODVGBAODV

Figure 16 Number of RERRs sent in highway scenario per connec-tion per second

model That is the reason why the reduction of the numberof RERRs in highway model is less than that in downtownmodel (referring to Figures 12 and 16)

To conclude GBAODV is much better than AODV inboth models It releases the load of the network (less numberof RREQs) reduces broken links and packet loss ratio andshortens average end-to-end delay

5 Vehicle Information Sinking NetworkBased on Mobile Nodes [35]

Aside of the mobile V2V network the information from thevehicles should also be sent to the sink node which will benormally performed by the roadside infrastructure Howeverthe construction of these infrastructure networks is expensivein both funding and time Hence mobile node acted by vehi-cles can firstly serve as the sinking port This section elabo-rates a data gathering algorithm based on swarm intelligenceAlthough the computational resource and energy sourceof the on-board computer in vehicles compared to fieldwireless sensor nodes is abundant applications may needto be extended to bicycle riders with limited energy source

0

003

006

009

012

50 60 70 80 90

Pack

ets l

oss r

atio

Number of nodes

Highway

AODVGBAODV

Figure 17 Packet loss ratio in highway scenario

0

20

40

60

80

100

120

50 60 70 80 90Number of nodes

Highway

AODVGBAODV

End-

to-e

nd d

elay

(ms)

Figure 18 Average end-to-end delay in highway scenario

The transmitted information in the future will be extended asthere will be applications aside of accident reporting such ascloud computationmedia access and entertainment throughthe V2V and V2I network Hence in order to maximizethe overall network efficiency communication load of eachvehicle node ought to be balanced

Much reference can bemade from the current research onhand-held devices such as 3Gmobile phone and PDA whichplay role as mobile sink of wireless sensor network (WSN)node in applications [36 37] Thus the algorithm for the datagathering application should support the sink node mobilityIt is a challenge in WSN algorithm design

Based on the application of the sensor network the datadelivery model to the sink node can be categorized intothree types query-driven event-driven and continuous Inquery-driven model the sink node generates a query andthen a temporary route is built The node which is checkedreceives query and returns result for instance DD [38] andACQUIRE [39] In event-driven model because the eventrate is much lower without temporal and spacial informationthe event node triggers the data transmission and temporaryroute building such as Rumor routing [40] and TTDD [41]Focusing on these two types of data transmission model theroutes are temporary so the sink node mobility has littleinfluence on data transmission In the continuous delivery

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 12: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

12 International Journal of Distributed Sensor Networks

model each sensor collects data periodically and sends datato the sink node for gathering In data gathering applicationthe sink node builds the route usually for instance it has beenconcluded in TEEN [42] APTEEN [43] and MINA [44]However movement of the sink node often results in brokenlinks If the route is rebuilt frequently not only the networkenergy consumption will be large but also the regular datatransmissionwill be blocked by the network stormwhichwillresult in massive broadcasting messages

Sink mobility brings new challenges in data gatheringapplication although some protocols and mechanisms havebeen proposed in recent years such as TDD SEAD [45]CODE [46] and others in [47 48] TTDD uses a gridstructure so that only sensor located at the grid points needsto acquire the forwarding information The route path forthe moving sink node is maintained and refreshed by agentnodes When there are several data sources in the networkthe overhead is largeMeanwhile the route needs to be rebuiltwhen the sink node moves out of the grid SEAD protocoldesigns a dormancy mechanism for the nodes in grid toreduce energy consumption The current route extends andrecovers by itself while the sink node moves but time delayremains a problem CODE does not need to rebuild globalpath but it needs other routing protocols to support sothe protocol is more complex A local routing restoringmechanism is proposed in [47] that the sink node sendsSink Claim message periodically This message is used forthe sink node detection The sensor node would changeits status according to this message The main problem ofthis method is the large consumption caused in sendingthe Sink Claim message in high frequency Meanwhile thethroughput becomes smaller As described in above TTDDSEAD CODE and others in [47 48] they are all designed forquery-driven or event-driven transmission model so thesemethods are notmuch suitable for data gathering application

For data gathering application in V2I this section studiesthe equilibriummechanism andproposes a swam intelligencedata gathering algorithm for mobile sink (SIDGMS) Theidea of SIDGMS algorithm is derived from swarm intelli-gence such as ants In this algorithm each vehicle node isa smart individual but with limited knowledge SIDGMSdefines two simple rules to describe the data forwardingTheproblem how to choose next hop becomes multiobjectiveprogramming which considers both the delay and load of thenetwork To solve link-break problem amethod of the powercontrol for the Sink beacon message is proposed

51 SIDGMS Algorithm The idea of SIDGMS algorithm isderived from swarm intelligence The preying behavior as atypical behavior of swarm intelligence has simple rules Ifan individual discovers food others will observe and studylocal behavior from the individuals in the region As a resulteveryone in group can find food

Each node in a wireless sensor network (WSN) systemhas limited computational ability memory space energy andwireless transmission range so the nodes can only exchangeinformation with neighbor nodes within the wireless com-munication range

Sink

Figure 19The structural relationship between the sink and vehicles

Sink node can be regarded as food source and the processof data gathering can be regarded as swarm foraging actionThe sink node and other nodes are mapped on Figure 19

The principle of data gathering works as below

(i) The sink node broadcasts beacon periodically whichcontains its current location information

(ii) The internal nodes discover the sink node directly andstart the data transmission with the sink node

(iii) Meanwhile the external nodes detect the data trans-mission between the internal nodes and the sinknode which helps the external nodes discover thesink indirectly and triggers the data transmissionbetween them if needed

(1) SIDGMS Algorithm Mechanism In this algorithm thenodes in WSN system can be separated into two types thesink node with mobility and the sensor nodes Two messagesare defined as below

Message 1 SINK BEACON (sinkInfo) which is sent by thesink node to inform sensor nodes sinkInfo includes thelocation component (119909 119910) and sequence number seq whichis incremental

Message 2 SENSOR DATA (nextAddr data sinkInfo load-Info) which is sent by sensor node The sensor node collectsits vehicular information (data) and then forwards it to next-hop node (nextAddr) The latest sink location component(sinkInfo) and load information component (loadInfo) areincluded in this message

During the data transmission period the sensor nodesaves location component of the sink node and refreshesthis component once it receives a new one which could beidentified by the sequence number component (seq) At thesame time the sensor node changes its status according to theSINK BEACONmessageThe status of sensor node is definedas follows

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 13: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of Distributed Sensor Networks 13

Definition 1 If sensor node receives the SINK BEACONmessage at a interval time 119879 (119879 is the periodic time ofSINK BEACON) the sensor node marks its status as SinkAdjacent (SA) Otherwise it marks its status as NonsinkAdjacent (NSA)

According to different status of the sensor node thealgorithmhas different data forwarding rules which are listedas follows

Rule 1 If the sensor node status is SA the data is forwardedto the sink node directly

Rule 2 For any sensor node in NSA status it has two criteriato choose next hopThe first criterion is for less delay and theother is for load balance of the network

Generally the sensor node which is closer to the sinknode has less jumping hops so its delay is smaller

For any sensor node 119894 the distance from the sink node iscalculated according to (4) as below

119889 (119894) = radic(119909119894minus 119909sink)

2

+ (119910119894minus 119910sink)

2

(4)

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the fast transmissionis solved according to (5)

119869 = arg min119895isin119873(119894)

119889 (119895) (5)

where119873(119894) denotes the neighbor nodes of the node 119894It is a complicated problem to calculate the loading of

each sensor node However it can be estimated in two waysIn one hand the key to maximize the WSN lifetime is toreduce energy consumption in each sensor node We assumethat each sensor node has the same hardware equipmentThus the remaining battery energy in each node should beconsidered

On the other hand the time to forward messages in aWSN system is closely related to the network performancesuch as the packet loss rate time delay and network conges-tion Therefore the total time of forwarding action could beused as indicator of the network loading It could be denotedby the number of package buffer queue

This mathematical model is elaborated as below

Definition 2 The loading of sensor node in WSN system is

119897 (119894) = 1198961 + 119902 (119894)

119890 (119894) (6)

where 119897(119894) denotes the loading of the sensor node 119894 119890(119894) is thebattery dump energy 119902(119894) is themean number of the packagesin the buffer queue and 119896 is scale factor

For any sensor node 119894 which is in NSA status the bestsensor node for next hop considering the load balance iscalculated by

119869 = arg min119895isin119873(119894)

119897 (119895) (7)

SinkJ

Figure 20 Same covering radius

Definition 3 For the criterions of Rule 2 120582119889is defined as

the distance weight coefficient with the sink node and 120582119897is

defined as the loading weight coefficient for the sensor node120582119889ge 0 120582

119897ge 0 and 120582

119889+ 120582119897= 1

The final optimization considers both the aspects of fastertransmission and power balance as below

min119895isin119873(119894)

(120582119889

10038161003816100381610038161003816119889 (119895) minus 119889

010038161003816100381610038161003816+ 120582119897

10038161003816100381610038161003816119897 (119895) minus 119897

010038161003816100381610038161003816)

st 119889 (119895) lt 119889 (119894) 120582119889ge 0

120582119897ge 0 120582

119889+ 120582119897= 1

(8)

where 1198890 is the minimum distance and 1198970 is minimumloading The distance weight coefficient 120582

119889and the loading

weight coefficient 120582119897are interrelated with application For

some applications which require high real-time response 120582119889

would be increased For some applications which focus onthe energy equilibrium 120582

119897will be increased 119889(119895) lt 119889(119894) to

prevent looping back

52 Power Control Strategy

(1) Node Coverage Radius The sink node broadcastsSINK BEACON periodically However no matter how fre-quently the sink node broadcasts the SINK BEACON mes-sage packet loss would happen That is because the linkagebetween the sink node and a sensor node in boundary areais fragile due to the movement of the sink node This case isshown in Figure 20 in which all the sensor nodes and sinknode have the same transmission radii

In order to solve this problem we propose a new strategyto avoid link breaking as shown in Figure 21 The transmis-sion radius of the sink node is smaller than that of the sensornodes At present most of the microcontrollers for WSN cansupport this by power control such as CC2430 CC1100 andMC132x

How to determine the transmission radius of theSINK BEACON As proposed in [49] with assumptions thatthe density of nodes is uniform and all nodes inWSNdomain

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 14: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

14 International Journal of Distributed Sensor Networks

SinkJ

Figure 21 Different covering radius

Sink

Rr

VT

Figure 22 Partial address routing

are subject to the Poisson distribution the probability that119898nodes exist in area 119878 is

119875 (119883 = 119898) =(120588119878)119898

119890minus120588119878

119898 (9)

Therefore the problemof radius 119903 could be translated into(10) as follows

119875 (119883 gt 0) = 1 minus 119875 (119883 = 0) ge 120572 (10)

where 120572 is the confidence which denotes the probability thatSA node appears

Finally the radius 119903 can be determined as

119903 ge radicln (1 minus 120572)

120587120588 (11)

(2) Cycle Time of SINK BEACON It is shown in Figure 22 Inthis case we assume the movement velocity of the sink nodeis 119881 the transmission radius of SENSOR DATA message is119877 and the transmission radius of SINK BEACONmessage is119903 (119903 lt 119877) Hence cycle time of the SINK BEACON messageshould satisfy this condition 119881119879 le 119877 minus 119903

Therefore the cycle time of the SINK BEACON is calcu-lated as follows

119879 le119877 minus 119903

119881 (12)

timerStart(119879 PERIOD TYPE)while (haveEnergy)

if (timerFired)sendMsg(SINK BEACON sinkInfo)

else if (receivedMsg)renderMsg()

endend

Pseudocode 1 The sink node pseudocode of SIDG-MS

status= NSAwhile (haveEnergy)

switch (receivedMsg)case SINK BEACON

status= SArecordSinkInfo()timerStart(119879 SINGLE TYPE)break

case SENSOR DATAif (toSelf)computeNextHop()forward()

elserecordInfo()

endbreak

endif (timerFired)status= NSA

endif (sensorDataReady)computeNextHop()sendData()

endend

Pseudocode 2 The sensor node pseudocode of SIDGMS

For example if the transmission radius of SEN-SOR DATA is 100m the transmission radius of SINK BEA-CON is 50m and the moving velocity of the Sink node is10ms representing that the sinking vehicle moves slowly onthe road the max cycle time of SINK BEACON is 5 s

53 Experiment and Analysis The simulation for evaluatingSIDG-MS algorithm is implemented with NS2

(1) Implementation of SIDG-MS Algorithm The sink nodetakes charge of sending SINK BEACON and data gatheringThe pseudocode is listed in Pseudocode 1

The sensor node collects vehicular data and forwards tothe sink It runs in distributed mode and the pseudocode inevery node listed in Pseudocode 2

(2) Test ScenarioThe simulation scenario is designed accord-ing to a plane area which is 800 meters wide and 800meters long There are totally 401 nodes in this WSN system

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 15: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of Distributed Sensor Networks 15

Table 6 Simulation parameters

Parameter ValueScene size 800 times 800 (m)Node number 400 node + 1 sinkMac 80211Application CBRPacket size 1024Queue length 10Channel model Two-ray ground

0 200 400 600 800 1000 1200

Alg

orith

m ty

pe

Time (hour)

Leersquos

Huangrsquos

This paper

Figure 23 Network lifetime comparison

including 1 sink node and 400 sensor nodes The sink nodemoves randomly in the network with a constant speed 10msThe sensor node collects the sensor data at a time intervalevery 10 s and its initialization energy is 50 JOther simulationparameters are listed in Table 6

We assume the energy consumption for collecting datais 1 times 10

minus5 J the energy consumption for receiving data is5 times 10minus5 J and the energy consumption for transmittingdata is 1 times 10minus4 J The value of SINK BEACON transmittingradius calculated according to (11) is larger than 50 meters(119903 ge 50m) in this simulation 119903 it is initialized as 80meter The transmission radius of the sensor node for SEN-SOR DATAmessage is initialized as 150mTheperiodic timeof SINK BEACON calculated according to (12) is smallerthan 7 s (119879 le 7 s) so 119879 is initialized as 5 seconds

(3) Simulation Result and Analysis The more incomingand outgoing message in MAC layer the larger energyconsumption will be Therefore we calculate the networkenergy consumption of every interval by counting the com-municationmessage inMAC layer isThe simulation result isshown in Figure 24 which illustrates the energy consumptioncharacter of SIDGMS Huang et al [47] and Lee et al [48]algorithms When the sink node is in motion the energyconsumption in literature [47] increases because the routepath increases In the algorithm of literature [48] the routepath to the sink node is checked during each message packettransmission so the energy consumption runs at a constantlyhigh level In SIDGMS algorithm the location of the sinknode could be refreshed during themotionThus this strategyhas less energy consumption

0 10 20 30 40 50 60 70 80 90 100 110 120 1300

10

20

30

40

50

60

70

80

140

LeersquosHuangrsquosThis paper

Ener

gy co

nsum

ptio

n (m

J)

Time (s)

Figure 24 The energy consumption comparison between threemethods

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14CBR sequence

Del

ay (m

s)

0

10

20

30

40

50

60

70

80

LeersquosHuangrsquosThis paper

Figure 25 Time delay comparison

The network lifetime is defined as the time period untilone of the nodes dies The simulation result with differentmethods is shown in Figure 23 The lifetime in literature [48]is the shortest due to the large energy consumption suchas many refreshing actions for route path The lifetime inliterature [47] is shorter than SIDGMS algorithm becausethere is no optimization for load balance Some nodersquosloadings are too heavy to support long lifetime

Time delay is an important factor ofWSN systemwe ana-lyze it bymonitoring the CBR streamThe simulation result isshown in Figure 25 Time delay with the SIDGMS algorithmis significantly lower than the other two algorithms

6 System Integration and Experiments [7]

To test the communication system we developed a series ofhardware as experimental platforms

61 Platform Integration Architecture Figure 26 showsthe system architecture including two major componentsonboard integration subsystem and V2V portable subsystem

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 16: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

16 International Journal of Distributed Sensor Networks

Onboard integration(full version) GUI

LCD

GPS signal

Driving behavior

GSMSMS

V2V wireless

PC104

GPS component

Vehicle sensor controller

GSM component

WLAN adapter

COM1COM2

COM3COM4

VGA

V2V modules(portable version) GUI

LCD

GPS signal

V2V wireless

PC104GPS component

WLAN adapter

COM1

COM4

VGA

middot middot middot

Figure 26 Platform integration architecture

156G accelerometer 500G accelerometer

Vehicle sensor controller

Acc pedal sensor Steering wheel sensor

WLAN subsystem(PC104 + GPS + GSM + WLAN adapter + GUI)

Test vehicle

Figure 27 Onboard integration subsystem

62 Integration for Onboard Subsystem Onboard subsystemis full version for collision detection and classification so allsensors as shown in Table 1 are installed onboard Somemainsensors are shown in Figure 27

63 Integration for V2V Portable Subsystem In order todesign low cost platform for V2V application we also needto develop a portable system to be installed on others test carA series of portableV2Vnodes have been developed and usedfor real road test as shown in Figure 28

Currently we implement GBAODV based on AODV-UU[50] Two threads are running under one main process One

is for routing in the network and the other is for reading GPSdata through serial port directly

Environment and devices for network test include(i) Linux Fedora 7(ii) PC104 consortium [51](iii) Ralink RT2500 series wireless LAN card(iv) SiRF StarIII GPS module(v) touch screen and keyboard

64 Road Test Scene In this section different experimentsare conducted to demonstrate the functions and performance

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 17: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of Distributed Sensor Networks 17

Figure 28 V2V portable subsystem

Figure 29 The scene of experiments Science Park Hong Kong

Table 7 Average packet loss ratio

Number of nodes AODV GBAODV4 455 6175 63 596 73 597 56 5438 515 41

of the integrated system In these experiments the key vehicleis a Toyota Corolla equipped with the full version systemincludingWLAN-based component GPS component GPRScomponent hazardous driving behavior detection subsys-tem and collision detection and analysis subsystem Asideof that we prepared eight sets of portable systems Theseportable systems include WLAN-based component and GPScomponent The scene of experiments is the road near HongKong Science Park and the corresponding driving path ismarked as a blue path in Figure 29

65 V2V Communication Test In this experiment (Figures30 and 31) all vehicles are driven along a line with 30 kmhrDifferent alarm signals are triggered manually by each of thevehicles randomly The source sends 100 PING messages todestination continuouslyThe V2V communication system isthen evaluated by checking whether the other vehicles canreceive the PINGmessage caused by status changingThe testresult is shown in Table 7

GBAODV performs better than AODV in generalAlthough the packet loss ratio is large this is acceptableSince there are barriers such as buildings in the experiment

Figure 30 Vehicle experiment

12

3

45

Figure 31 GUI for vehicle experiment

environment the signal attenuates rapidly The packet lossratio after one hop is approximately 20 PING is roundtrip message If source and destination cannot communicatedirectly PING message traverses at least 4 hops Thereforethe packet loss ratio is at least

1 minus (1 minus 02)4

= 05904 (13)

This is close to the experiment results If the environmentis clear enough the results should be better

7 Conclusion

In this paper we presented a vehicle safety enhancementsystem based on wireless communication The system canobtain vehicular signals classify hazardous information andmake decision to trigger different actions to prevent theaccident from occurrence or deterioration To enhance thenetwork performance we evaluated DSDV DSR and AODVprotocols and adopted AODV as the benchmark protocolThereafter GPS information is integrated into AODV tofurther upgrade to GBAODV which reduces packet loss rateand end-to-end delay especially for downtown application inVANETThis paper also addresses V2I routing by proposingthe SIDGMS which balances delay and network load Sim-ulation validates the V2I algorithm Finally we evaluate theV2V system by on-road test

Acknowledgments

The authors would like to Dr Xin Shi Dr Wing KwongChung Mr Yanbo Tao Mr Kai Wing Hou Mr MaxwellChow for participating in the project and the on-roadtest This paper is partially supported by the Hong KongInnovation and Technology Fund project ITP00309AP and

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 18: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

18 International Journal of Distributed Sensor Networks

RFD20122013 at Centre for Research onRobotics and Smart-city The Chinese University of Hong KongmdashSmart China

References

[1] httpwwwvehicle-infrastructureorg[2] M Heddebaut J Rioult J P Ghys C Gransart and S Ambel-

louis ldquoBroadband vehicle-to-vehicle communication using anextended autonomous cruise control sensorrdquo MeasurementScience and Technology vol 16 no 6 pp 1363ndash1373 2005

[3] M Shulman and R Deering ldquoThird annual report of the crashavoidance metrics partnership April 2003-March 2004rdquo TechRep DOTHS 809 837 National Highway Traffic Safety Admin-istration Washington DC USA 2005

[4] W Franz R Eberhardt and T Luckenbach ldquoFleetnetmdashinterneton the roadrdquo in Proceedings of the 8th World Congress onIntelligent Transport Systems Sydney Australia 2001

[5] T Taleb E Sakhaee A Jamalipour K Hashimoto N Kato andY Nemoto ldquoA stable routing protocol to support ITS services inVANET networksrdquo IEEE Transactions on Vehicular Technologyvol 56 no 6 pp 3337ndash3347 2007

[6] OnStar httpwwwonstarcom[7] Y Chen Y Sun N Ding et al ldquoA real-time vehicle safety sys-

temrdquo in Proceedings of the IEEESICE International Symposiumon System Integration pp 957ndash962 Fukuoka Japan December2012

[8] httpenwikipediaorgwikiTraffic collision[9] K LaMance ldquoAuto accidents caused bysudden stopsrdquo http

wwwlegalmatchcomlaw-libraryarticleauto-accidents-caus-edby-sudden-stopshtml

[10] Krystle B Robertson N Shaw and et al ldquoHow to deal witha minor car accidentrdquo httpwwwwikihowcomDeal-With-a-Minor-Car-Accident

[11] L A Wallis and I Greaves ldquoInjuries associated with airbagdeploymentrdquo Emergency Medicine Journal vol 19 no 6 pp490ndash493 2002

[12] httpenwikipediaorgwikiRollover[13] N LiuHQian J Yan andY Xu ldquoPerformance analysis of rout-

ing protocols for vehicle safety communications on the freewayrdquoin Proceedings of the 3rd International Conference on Anti-Counterfeiting Security and Identification in Communication(ASID rsquo09) pp 85ndash88 Hong Kong August 2009

[14] M Zhang and R S Wolff ldquoRouting protocols for vehicular AdHoc networks in rural areasrdquo IEEE Communications Magazinevol 46 no 11 pp 126ndash131 2008

[15] S-J Lee M Gerla and C K Toh ldquoSimulation study of table-driven and on-demand routing protocols for mobile ad hocnetworksrdquo IEEE Network vol 13 no 4 pp 48ndash54 1999

[16] A Laouiti P Muhlethaler F Sayah and Y Toor ldquoQuantitativeevaluation of the cost of routing protocol OLSR in a vehicle adhoc network (VANET)rdquo in Proceedings of the IEEE Conferenceon Vehicular Technology pp 2986ndash2990 Singapore May 2008

[17] L Abusalah A Khokhar and M Guizani ldquoA survey of securemobile ad hoc routing protocolsrdquo IEEE Communications Sur-veys and Tutorials vol 10 no 4 pp 78ndash93 2008

[18] IETF MANETWorking Group httpwwwietforghtmlchar-tersmanetcharterhtml

[19] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on Communications

Architectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 London UK September 1994

[20] D Johnson Y Hu and D Maltz ldquoThe dynamic source rout-ing protocol (DSR) for mobile ad hoc networks for IPv4rdquoIETF MANET Working Group Internet Draft (RFC4728)httpwwwietforgrfcrfc4728txt

[21] C Perkins E Belding-Royer and S Das ldquoAd hoc on-demanddistance vector (AODV) routingrdquo IETF MANET Work-ing Group Internet Draft (RFC3561) httpwwwietforgrfcrfc3561txt

[22] Y Sun Y Chen and Y Xu ldquoA GPS enhanced routing protocolfor vehicular ad-hoc networkrdquo in Proceedings of the IEEE Inter-national Conference on Robotics and Biomimetics pp 2096ndash2101 Phuket Thailand December 2011

[23] C E Perkins and E M Royer ldquoAd-hoc on-demand distancevector routingrdquo in Proceedings of the 2nd IEEE Workshop onMobile Computing Systems and Applications (WMCSA rsquo99) pp90ndash100 New Orleans La USA February 1999

[24] WXiong andQ Li ldquoPerformance evaluation of data dissemina-tions for vehicular ad-hoc networks in highway scenariosrdquoTheInternational Archives of the Photogrammetry Remote Sensingand Spatial Information Sciences vol 37 part B1 2008

[25] B Mustafa and U W Raja Issues of routing in VANET [MSthesis] School of Computing Blekinge Institute of TechnologyBlekinge Sweden June 2010

[26] W Kim D Kwon and Y Suh ldquoA reliable route selectionalgorithm using global positioning systems in mobile Ad-hocnetworksrdquo in Proceedings of the IEEE International ConferenceonCommunications pp 3191ndash3195Helsinki Finland June 2001

[27] O Abedi R Barangi and M A Azgomi ldquoImproving routestability and overhead of the AODV routing protocol andmaking it usable for VANETsrdquo in Proceedings of the 29th IEEEInternational Conference on Distributed Computing SystemsWorkshops Montreal Quebec Canada June 2009

[28] O Abedi M Fathy and J Taghiloo ldquoEnhancing AODV routingprotocol using mobility parameters in VANETrdquo in Proceedingsof the IEEEACS International Conference on Computer Systemsand Applications pp 229ndash235 Doha Qatar March-April 2008

[29] H Asenov and V Hnatyshin ldquoGPS-enhanced AODV routingrdquoin Proceedings of the International Conference on WirelessNetworks (ICWN rsquo09) Las Vegas Nev USA 2009

[30] VanetMobiSim httpvaneteurecomfr[31] NS2 httpwwwisiedunsnamns[32] TIGERLinecircledR httpwwwcensusgovgeowwwtiger[33] M Treiber A Hennecke and D Helbing ldquoCongested traffic

states in empirical observations and microscopic simulationsrdquoPhysical Review E vol 62 no 2 pp 1805ndash1824 2000

[34] ldquoVanetMobiSim Manualrdquo httpwwwscribdcomdoc1107-52020VanetMobiSim-1-0-Manual

[35] Y Chen Y Tang G Xu H Qian and Y Xu ldquoA data gatheringalgorithm based on swarm intelligence and load balancingstrategy for mobile sinkrdquo in Proceedings of the 8th WorldCongress on Intelligent Control and Automation (WCICA rsquo11)pp 1002ndash1007 Taipei Taiwan June 2011

[36] K Akkaya and M Younis ldquoA survey on routing protocols forwireless sensor networksrdquo Ad Hoc Networks vol 3 no 2 pp325ndash349 2005

[37] I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci ldquoAsurvey on sensor networksrdquoAdHoc Networks vol 40 no 8 pp102ndash105 2002

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 19: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of Distributed Sensor Networks 19

[38] C Intanagonwiwat R Govindan and D Estrin ldquoDirected dif-fusion a scalable and robust communication paradigm forsensor networksrdquo in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOMrsquo00) pp 56ndash67 Boston Mass USA August 2000

[39] N Sadagopan B Krishnamachari and A Helmy ldquoTheACQUIRE mechanism for efficient querying in sensor net-worksrdquo in Proceedings of the IEEE International Workshop onSensor Network Protocols and Applications pp 149ndash155 2003

[40] D Braginsky and D Estrin ldquoRumor routing algorithm forsensor networksrdquo in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNArsquo02) pp 22ndash31 Atlanta Ga USA September 2002

[41] H Luo F Ye J Cheng S Lu and L Zhang ldquoTTDD two-tier data dissemination in large-scale wireless sensor networksrdquoWireless Networks vol 11 no 1-2 pp 161ndash175 2005

[42] AManjeshwar andD PAgrawal ldquoTEEN a routing protocol forenhanced efficiency in wireless sensor networksrdquo in Proceedingsof the 15th International Parallel and Distributed ProcessingSymposium pp 2009ndash2015 San Francisco Calif USA 2001

[43] A Manjeshwar and D P Agrawal ldquoAPTEEN a hybrid protocolfor efficient routing and comprehensive information retrievalin wireless sensor networksrdquo in Proceedings of the 15th Interna-tional Parallel and Distributed Processing Symposium 2002

[44] J Ding K Sivalingam R Kashyapa and L J Chuan ldquoA multi-layered architecture and protocols for large-scale wireless sen-sor networksrdquo in Proceedings of the IEEE Vehicular TechnologyConference (VTC rsquo03) vol 3 pp 1443ndash1447 Orlando Fla USAOctober 2003

[45] Y C Hu D B Johnson and A Perrig ldquoSEAD secure efficientdistance vector routing for mobile wireless ad hoc networksrdquoAd Hoc Networks vol 1 no 1 pp 175ndash192 2003

[46] L X Hung D H Seo S Lee and Y K Lee ldquoMinimum-energydata dissemination in coordination-based sensor networksrdquoin Proceedings of the 11th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applicationspp 381ndash386 Hong Kong August 2005

[47] Q Huang Y Bai and L Chen ldquoAn efficient route maintenancescheme for wireless sensor network with mobile sinkrdquo inProceedings of the IEEE 65th Vehicular Technology Conference(VTC rsquo07) pp 155ndash159 Dublin Irland April 2007

[48] D Lee S Park E Lee Y Choi and S-H Kim ldquoContinuousdata dissemination protocol supporting mobile sinks with asink location managerrdquo in Proceedings of the Asia-Pacific Con-ference on Communications (APCC rsquo07) pp 299ndash302 BangkokThailand October 2007

[49] T S RappaportWireless Communications Principles and Prac-tice IEEE Press Piscataway NJ USA 1996

[50] AODV-UU httpcoreituuseadhoc[51] PC104 Embedded Consortium httpwwwpc104orghistory

php

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 20: Research Article Vehicle Safety Enhancement System ...downloads.hindawi.com/journals/ijdsn/2013/542891.pdf · Research Article Vehicle Safety Enhancement System: ... in the aspects

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of


Recommended