Research ArticleResearch on Joint Handoff Algorithm in Vehicles Networks
Yuming Bi Lei Tian Mengmeng Liu Zhenzi Liu and Wei Chen
Key Laboratory of Universal Wireless Communications Ministry of Education Beijing University of Posts and TelecommunicationsMailbox No 92 Beijing 100876 China
Correspondence should be addressed to Yuming Bi biyuming10507sinacom
Received 1 October 2015 Revised 12 January 2016 Accepted 21 March 2016
Academic Editor Sujit Dey
Copyright copy 2016 Yuming Bi 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 communication services evolution from the fourth generation (4G) to the fifth generation (5G) we are going to facediverse challenges from the new network systems On the one hand seamless handoff is expected to integrate universal accessamong various network mechanisms On the other hand a variety of 5G technologies will complement each other to provideubiquitous high speed wireless connectivity Because the current wireless network cannot support the handoff among WirelessAccess for Vehicular Environment (WAVE) WiMAX and LTE flexibly the paper provides an advanced handoff algorithm to solvethis problem Firstly the received signal strength is classified and the vehicle speed and data rate under different channel conditionsare optimized Then the optimal network is selected for handoff Simulation results show that the proposed algorithm can welladapt to high speed environment guarantee flexible and reasonable vehicles access to a variety of networks and prevent ping-ponghandoff and link access failure effectively
1 Introduction
With the continuous development of the fifth generation (5G)mobile communications and the ever-maturing standardsintelligent terminals and mobile Internet and networkingsystems have shown a trend of rapid development worldwideMeanwhile the vehicular communication system has a widerange of engineering application values in the future commu-nication markets Therefore strong research activities havebeen stimulated in the field of Wireless Access for VehicularEnvironment (WAVE) However due to the high mobilityof vehicles the seamless handoff becomes a new bottleneckin vehicular communication system Several literatures haveinvestigated the handoff problem Reference [1] proposed ahandoff method based on the detection of received signalstrength (RSS) while [2] mainly focused on the traditionaldata rate In [3] the packet loss rate was considered todecide handoff In [1ndash3] certain physical quantities alsocalled detection values are utilized to judge whether thehandoff should be done Besides the cost functions canalso be used to make handoff decisions such as throughputand quality of service (QoS) [4ndash6] Some scholars believedthat multiple factors should be considered Thus several
handoff algorithms based on multiple attribute decision(MAD) appeared The operator policies terminal propertiescustomer performance and the application QoS level are alltaken into account [7 8] In addition [9] used an analytichierarchy process (AHP) to acquire the various performanceparameters to make a judgment for the weights of thenetwork Reference [10] utilized the remaining bandwidthof vertical handoff algorithm for switching among hetero-geneous networks relying on the QoS Basic Service Set(QBSS) with limitation All of the above-mentioned handoffalgorithms aimed at cellular communication systems Thereis still lack of research on handoff problem in vehicularcommunication system Different from the traditional cel-lular networks which are based on fixed base stations (BSs)and mobile cellular networks the vehicular communicationsystems require high speed mobile terminals (MTs) to useself-organizing network (SoN) or cooperation to realize end-to-end (e2e) data interaction directly [10] As a consequencethe vehicular communication systemmay easily produce calldropping and ping-pong effects In addition some papersfocused on the vertical handoff and the others only tookthe horizontal handoff into consideration But it is clear thatthe future wireless access network will consist of wireless
Hindawi Publishing CorporationChinese Journal of EngineeringVolume 2016 Article ID 3190264 10 pageshttpdxdoiorg10115520163190264
2 Chinese Journal of Engineering
networks with various services and coverage ranges So asto accurately obtain the ldquoalways best connectrdquo performancenetworks the joint vertical and horizontal handoff in het-erogeneous networks will become an increasingly importantresearch hotspot although little researche has been done onit In order to allow users to roam among various systemstransparently and seamlessly this paper presents a novelalgorithm which not only supports the horizontal handoffin the same network but also can classify the RSS andoptimize velocity speed and data rate under different channelconditions to choose an optimal network to handoff Forthe sake of low computational complexity RSS is adoptedto handle horizontal handoff and the fuzzy logic decisionalgorithm is used to address vertical handoff There is nodoubt that the fuzzy logic decision algorithm takes many fea-tures of the mobile terminals (MTs) and access network intoconsideration and thus makes the handoff algorithm easierfor hardware implementation The remainder of this paperis organized as follows Section 2 describes the backgroundof handoff process Section 3 indicates the proposed handoffalgorithm and gives the related channel model Numericalresults are presented in Section 4 Finally Section 5 concludesthe whole paper
2 Background
21 The Brief of Handoff It is not difficult to find that the5G wireless access network will consist of access networkswith various coverage and data rates The hybrid networksystem can take advantage of each system to provide highQoS to users According to the relationship between thesource and target network the handoff process can be dividedinto horizontal handoff and vertical handoff as shown inFigure 1 In general horizontal handoff refers to handoffin the same access technology while vertical handoff is thehandoff process occurring between different access networktechnologies No matter which handoff process happens theultimate objective is to keep ldquoalways best connectrdquo (ABC) forusers [11ndash13] That is to say when a mobile user connectsto hybrid access networks concurrently all they will dois to choose a best adjacent cell in the same network oraccess network for its service requirement purpose Inmobilecommunication systems there exists a certain overlap areabetween adjacent cells called handoff belt When the mobileterminal moves into this area the wireless link handoffprocessmust be conducted effectively However in the hybridnetwork systems the coverage area of several networks maybe overlapped Thus a new handoff scenario in the hybridnetwork system has to occur in the near future [14] In orderto minimize the link dropping probability the horizontal andvertical handoff should be considered jointly
22 The Network Architecture and Protocol The existingnetwork architecture consists of mobile terminal accessnetwork and core network In the access network thebase station (eNodeB) bears the control function of wirelesssystem and makes the data transmission rate become fasterMoreover the boundaries between access network and corenetwork are blurredThemobile terminal is connected to the
Horizontalhandoff
LTE
Horizontalhandoff
Horizontalhandoff
WLAN
WAVE
Verticalhandoff
Verticalhandoff
Figure 1 The schematic map of handoff
LTE3G 2G
Corenetwork
Accessnetwork
S1 S1
X2eNodeB
eNodeB
Figure 2 The vehicular wireless communication system
access network via air interfaceThe eNodeB uses S1 interfaceto connect to the core network and X2 interface to link withanother eNodeB respectively as shown in Figure 2
The typical vehicular communication system is dividedinto three subsystems core network vehicle and groundparts The vehicular subsystem mainly includes the vehicu-lar information platform the communication module thesystem multiplexing module the 2G3G automotive microbase station unit and theWi-Fi hotspots equipment It alwaysdirectly communicates with the ground stations throughstandard onboard processing unit which is responsible forthe large capacity data transmission The ground subsystemmainly includes the eNodeB radio remote unit (RRU) andthe S1 and X2 interface optic fibers This paper assumesthat the parts of the core network are related to the accessnetwork Specifically the core network uploads the users
Chinese Journal of Engineering 3
service information and the whole system control informa-tion Moreover the access network is utilized for convertingthe format of 2G3G and separating them into the differentservice networks
Therefore as discussed above there are two types ofhandoff process one is based on X2 interface switch andthe other is connected with S1 interface For the horizontalhandoff the data will be passed throughX2 interface betweeneNodeBs When the user equipment (UE) needs verticalhandoff the source eNodeB will trigger the S1 interface toswitch information from eNodeB to access network and corenetwork
23 The Procedure of Handoff The handoff process has threesteps (1) Measurement of handoff at first UE always triesto find out whether there is a new available network ornot In the vertical handoff process many conditions ofswitch are considered such as signal receiving intensity userpreference network cost and load balance The commonhandoff decision will be triggered when network congestionis serious or current network signal and QoS are weak Thehandoff may also be held when the current network cannotprovide satisfactory businesses to the user (2) Switchingdecision stage there are three kinds of control modes tomake the handoffdecision network control terminal controland terminal auxiliary switch The system will syntheticallytake into account the situation of each alternative networkthe terminal characteristics and the userrsquos preferences forthe current business In some special circumstances otherfactors (such as QoS and system performance security) willalso be considered (3) The execution phase of handoff thisstage is completing the handoff process and switching thecommunication services from the current access point to thetarget network To ensure the completion of the handoffsome protocols are neededmore details about it can be foundin [15]
3 The Handoff Algorithm
In this section we present a novel scheme based on thejoint RSS and fuzzy logic handoff algorithm Figure 3 showsthe algorithm flow chart Assume that the WLAN systemhas successfully established the connection The vehiclemeasurement equipment always detects the various accessnetwork signals Once the RSS is higher than that from otheraccess networks the station keeps the data connection withWLAN and meanwhile rises the horizontal handoff modeOtherwise the mobile station waits for the optimal networkselection time before comparing the RSS again After the RSScomparison if the RSS in WLAN is still higher than theothers the whole handoff process returns to the initial phaseand the vehicle measurement equipment retests the variousaccess network signals Otherwise the algorithmmechanismwill decide to start the vertical handoff mode which makesdecision based on fuzzy inference engine by utilizing Zadehrules and the details are described as follows
31 The Algorithm of Horizontal Handoff During eachperiod the UE in either idle or connection state measures
Select the target cell with the same networkaccording to the processof UE moving
Dynamically adjust the sampling periods duringmeasurement based on the received signal
Horizontal handoffswitching decision
Start the verticalhandoff mode
Measure the relatedparameters ofvarious networkaround the UE
Fuzzy inferenceengine based onZadehrsquos method
Vertical handoffswitching decision
YesNo
Yes
No
Initial phase(connect with the source access network)
Real‐time detection of all kinds of signalsstrength from various networks
Waiting a dwelltime Then ifSsource lt Sselection
handoff modeStart the horizontal
Ssource lt SselectionIf
Figure 3 The flow chart of handoff algorithm
RSSI
Fuzzy inferenceengine based onZadehrsquos method
Data rate Knowledgerules
Reversalfuzzifier
APCVoutput
Interferenceratio
Figure 4 The fuzzy logic decision scheme
the related power parameters related to both the source celland the target one The measurement parameters includereference signal received power (RSRP) reference signalreceived quality (RSRQ) and reference signal strength indi-cator (RSSI) The RSRP is the received power value of thepilot signal from base station It is calculated by the difference
4 Chinese Journal of Engineering
between the reference signal which transmits from the basestation and the path loss
RSRP = 119875119879119909minus PL (dB) (1)
where 119875119879119909
is the transmitter power of the pilot signal and PLrepresents the path loss The RSRQ can be calculated as
RSRQ = 119873 lowast RSRPRSSI
(2)
where 119873 is the number of Resource Blocks (RB) in termsof bandwidth The RSSI is the measurement value of thereceived carrierrsquos power in thewhole systembandwidth how-ever the carrier sends not only data and control informationbut also some interference and noise information In orderto make the handoff decision reasonable the acquired infor-mation from the measurement data needs to be submittedto the Radio Resource Control (RRC) layer In additionthe measured reference signal should be smoothed by filtersbefore being transmitted to the RRC for eliminating randomfluctuations [16] The trigger condition of handoff executionis as follows the QoS of target cell is better than the currentlyserving cell and the difference exceeds a specified thresholdwhen the duration is greater than the trigger delay time Thejudgment criteria are shown as
where 119875target and 119875source represent the strength power of thetarget cell and source cell respectively and 119874
119891means the
offset between related source cell and target cell When themeasurement value of the received signal is consistent with(3) the timer will start working If the relationship describedby (3) continuity satisfies the rule of time to trigger (TTT) thesystemwill determine to drive the handoff process executionThe decision threshold 119867 and TTT have great influenceon the performance of the algorithm Larger 119867 and longerdelay time of departure will make the handoff more difficultTherefore as a result of switch it is easy to lead to linkconnection fail namely drop link If 119867 is too small it willcause frequent switching namely ping-pong phenomenon
32 The Algorithm of Vertical Handoff In this part we willfocus on how to use the algorithm for a vertical operationbased on fuzzy logic As we all know the general elementsof collection set for the membership can only take 0 and 1 In1965 L A Zadeh expanded the membership from only twovalues (0 and 1) to any values within 0 and 1 which was asign that the membership function with fuzzy sets could berepresented by fuzzy probability The fuzzy subset 119861 whichbelongs to the theory domain 119879 is a function characterizedby a collection of membership and can be mapped as follows
120583119861 119879 997888rarr [0 1] (4)
where120583119861means themembership function of the fuzzy subset
120583119861(119905) represents the extent to which the elements 119905 in the
theory domain119879 belong to fuzzy subset 119861The larger value ofit the higher probability of belonging to 119861 For a given theory
TargetBSSource BS
y-axis
x-axis(0 0)
(0 50)(2000 50)
(2000 0)
Figure 5 The horizontal handoff simulation model
LTE-eNodeBWLAN-BS
y-axis
x-axis(0 0) (500 0)
(0 minus50) (250 minus50)
(2200 0)
Figure 6 The vertical handoff simulation model
domain of 119864 a kind of word which is related to 119864 constitutesa set 119875 Its semantics is represented by the function 119877 whichmaps the relationship between set 119875 and 119864 Therefore theword set is a fuzzy function and it can be described as
where 119890 is the element of 119864 And the membership function120583119861(119886 119890) means the extended relationship between 119886 (which
belongs to the set 119875) and 119890 (which belongs to the theorydomain 119864)
120583119877 119875 times 119864 997888rarr [0 1] (6)
In order to deduce the fuzzy logic relationship we willintroduce the principle of Zadehrsquos method if the fuzzyrelationship ldquoIf 119860 then 119861rdquo can be represented by ldquo119860 rarr 119861rdquowhere119860 isin 119880 119861 isin 119881 then the fuzzy logic relationship119877(119906 V)is defined as
Zadeh119877 (119906 V) = (119860 (119906) and 119861 (V)) or (1 minus 119860 (119906)) (7)
where ldquoandrdquo and ldquoorrdquo stand for the supremum and infimumoperatorwith logic relationship respectively For a given119860lowast isin119880 if the relationship matrix 119877 is known then 119861lowast isin 119881 can becalculated by
119861
lowast= 119860
lowast∘ 119877 (8)
where ldquo∘rdquo means the supremum operator with logic relation-ship
The proposed fuzzy logic decision strategy consists ofthree steps (as shown in Figure 4) fuzzy inference enginebased on Zadehrsquos method reversal fuzzifier and APCVoutput At first the input parameters such as RSSI IR andDR are mapped into inference engine by utilizing knowledge
Chinese Journal of Engineering 5
0 500 1000 1500 2000 2500 30000
200
400
600
800
1000
1200
1400
1600
1800
2000
The radius of cell (m)
Aver
age d
wel
l tim
e (s)
5kmh30kmh60kmh
75kmh90kmh120 kmh
(a)
0 2000 4000 6000 8000 100000
01
02
03
04
05
06
07
08
09
1
The radius of cell (m)
Prob
abili
ty o
f han
doff
5kmh30kmh60kmh
75kmh90kmh120 kmh
(b)
Figure 7 The handoff probability and the average time with various cell radius and user speeds
rules And then the inference engine will make a fuzzydecision according to the mapping values But as mentionedabove the output of the fuzzy inference module is still afuzzy set and has no effects on the controlled object directlyTherefore a ldquotranslationrdquo process is needed to transfer a fuzzyvalue to a precise one119891
0This step is called ldquoreversal fuzzifierrdquo
Through comparison with this precise value we can get theaccess point candidacy value (APCV) output The goal ofthis ldquoreversal fuzzifierrdquo is to find the maximum membershipdegree function The ldquoreversal fuzzifierrdquo selects the elementwhich will make the value of membership function thehighest as the output signal And the rules could be describedas
120583 (1198910) ge 120583 (119891) 119891 isin 119865 (9)
In case of nonunique maximum value of the membershipfunction the average value will be taken as follows
1198910=
1
119873
119873
sum
119894=1
119891119894 120583 (119891
119894) ge 120583 (119891) (10)
In different networks different fuzzy inference modelsare adopted The membership function is designed to ensurecomparability among different networks at the same timeAt last in order to complete the behavior of handoff thesystem will select the highest APCV network to accessThereare three parameters that need to be detected in the fuzzymodule RSSI data rate and interference ratio Gaussian and119878-scheme distribution are chosen as the fuzzy membershipbecause of their great performance with respect to real-timecontrol
Figure 9 shows the WLAN membership functions Forexample RSSI stands for the received signal strength indi-cator Therefore we make such an assumption to describe its
knowledge rules (as shown in Figure 9(c)) (i) the member-ship function is 119878-scheme distributionwhenRSSI is in a giveninterval from minus95 dBm to minus80 dBm (ii) if RSSI belongs to theinterval [minus90 dBmminus70 dBm] then themembership functionfollows Gaussian distribution and (iii) once the value ofRSSI is higher than minus80 dBm the membership function is119878-scheme distribution as well The fuzzy set of RSSI the setof data rate and the fuzzy set of interference ratio in WLANare [good normal weak] with value set [3 2 1] Hence theoutput fuzzy sets are mapped as
119878 =
3
sum
119894=1
119872119894119896 (11)
Take a regular WLAN network as an example if RSSI isweak and data rate is weak and interference ratio is weakthen 119878 is 3 For the WLAN network each input has threefuzzy sequences so there are 27 fuzzy rules in the knowledgerules Due to the sensitivity to UE speed in V2V network5 fuzzy variable sequences are used (45 fuzzy rules) in theknowledge rules The input parameters in the fuzzy logicmodule are mapped into different fuzzy sets and the outputwould be acquired utilizing the maximum membershipdegree function method The fuzzy logic outputs decisionvalue indicates that the stable degree of difference betweentwo network decisions is lower than the threshold level thusthe connection with the source network cell will be held onOtherwise the handoff executionwill be done if the hysteresistime is longer than a certain value that is
VALUE TARGET minus VALUE SOURCE ge Hysteresis
VALUE SOURCE le Threshold(12)
6 Chinese Journal of Engineering
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The s
treng
th o
f rec
eive
d sig
nal
from
targ
et ce
ll (d
Bm)
The s
treng
th o
f rec
eive
d sig
nal
from
sour
ce ce
ll (d
Bm)
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The location of vehicle (km)
The location of vehicle (km)
(a) The strength of received signal from source and target cell
06 07 08 09 1 11 12 13 140
01
02
03
04
05
06
07
08
09
1
The location of vehicle (km)
The p
roba
bilit
y of
han
doff
The theoretical handoff point
The probability of connectingwith the target cell
The probability of connectingwith the source cell
(b) The probability of handoff
06 07 08 09 1 11 12 13 1401
02
03
04
05
06
07
08
09
10
The location of vehicle (km)
Prob
abili
ty o
f han
doff The actual handoff location point
(c) The actual handoff location point
20 40 60 80 100 120 140 160 1800
005
01
015
02
025
03
035
04
The speed of mobile terminal (kmh)
Ping
-pon
g ha
ndoff
ratio
Proposed algorithmExisting RSSI algorithm(d) The performance comparison
Figure 8 The horizontal handoff simulation results
33 Channel Model This subsection will adopt the COST 231Hata channel model to realize the real-time simulation [1718] which is depicted as
PL (dB) = (397 minus 701 log10 (ℎBS)) log10 (119889
1000
)
+ 429 + (3654 minus 12ℎMS) log10 (119891119888)
minus 1298 log10(ℎBS) + 074ℎMS + 119862
(13)
where ℎBS and ℎMS represent the height of BS and MSrespectively 119889 means the distance between BS and MS 119891119888denotes the center frequency and 119862 is a constant
The channel impulse response (CIR) from transmitterantenna element 119904 to receiver element 119906 for cluster 119899 isexpressed as
is the cross polarization power ratio in linear scale V is themoving speed of user and 120579
119899119898AoA and 120579119899119898AoD mean the
angle of arrival (AoA) and angle of departure (AoD) with the119898th subpath in 119899th path respectively Assume that there are6 paths and each path includes 20 subpaths
4 The Simulation and Validation
To evaluate the realistic performance of our proposed algo-rithm the simulation model scenarios and various parame-ters are developed using MATLAB The performance of hor-izontal and vertical handoff algorithms is tested separately
Figure 7 compares the handoff probability and averagedwell time with different moving speed and cellular radius Itis found that the expansion of cellular radius and decreasingvehicle moving speed will lead to increase of average dwelltime and channel holding time In particular the relationshipbetween the coverage radius and dwell time is obviouslylinear in the low vehicle speed regions On the contrary withthe increase of vehicle speed the handoff probability willcontinue to decrease
The horizontal handoff is simulated using the followingparameters (as shown in Figure 5) the point coordinates of
source BS and target BS are [0 50] and [2000 50] whichmeans that the distance between the source BS and targetBS is 2000m and the distance between road and BS is50m the vehicle speed is 80 kmh and is moving from [0 0]to [2000 0] the heights of BS and MS are 30m and 1mseparately the transmitted power is 44 dBm and the standarddeviation of channel shadow fading is 8 dB
Figure 8 shows the simulation results of horizontal hand-off It is clearly seen that when the vehiclemoves from [500 0]to [1500 0] the strength of received signal from source BSis reducing while the signal level from target BS is growingMeanwhile the handoff probability shows a similar trendand the handoff should occur at the points near [1000 0]according to the theoretical judgment of holding probabilityIt can also be observed that the handoff happens at the peakof the curve namely the coordinate of the vehicle is [1055 0]There is a little far distance from the actual handoff locationto the theoretical one due to the hysteresis effect of dwelltime on the joint handoff algorithm As we all know theexisting RSSI algorithm depends on the decisions of priorityon the mobile node In particular the priority is dividedinto low priority and high priority The low priority whichis below certain level estimate receives signal strength from
8 Chinese Journal of Engineering
0 50 100 150 200 250 300 350 400minus105
minus100
minus95
minus90
minus85
minus80
minus75
minus70
The simulation time (s)
RSS
(dBm
)The received signal strength
LTEWLAN
(a) RSS
0 50 100 150 200 250 3000
01
02
03
04
05
06
07
08
09
1
The simulation time (s)
Mar
k of
net
wor
k
LTE
WLANThe state of handoff
(b) The state of handoff
50 100 150 200 250 300 3500
1
2
3
4
5
6
7
8
9
10
The simulation time (s)
APC
V
LTEWLANV2V
V2VWLAN
LTE
(c) APCV and state of handoff
1 15 2 25 3 35 402
03
04
05
06
07
08
09
1
Request (MBs)
Thro
ughp
ut
V2VWLANLTE
(d) Throughput versus request
Figure 10 The vertical handoff simulation results
another cell before it starts handoff in the cell ThereforeFigure 8(d) presents a performance comparison between theexisting RSSI horizontal handoff algorithm and our proposedschemes It is clear that no matter how much the speedof users is the ping-pong handoff ratio of existing RSSIhorizontal handoff algorithm is higher than the proposedschemes It proves that ping-pong effect can be avoidedeffectively by using dwell timer and hysteresis parameters
The vertical handoff simulation is divided into two partsone focuses on testing whether the handoff could be appliedautomatically and another part analyzes the performanceof the proposed algorithm Firstly assume a car is movingfrom the initial points [0 minus50] to [2500 minus50] at the speed of45 kmh and the coordinates of WLAN-BS and LTE-eNodeare [500 0] and [2200 0] respectively (as shown in Figure 6)
Figure 10(a) shows the alternative trend of the RSS during thewhole movement and Figure 10(b) shows the actual handoffresults It is not difficult to find such a handoff process thecar accessed LTE at first and then changed the network toWLAN because it ran close to theWLAN-BS when it left thecoverage of WLAN the handoff happened again to replaceWLAN by LTE signals Obviously because of different pathloss and shadow fading due to different center frequencythe vehicle utilizes LTE network to communicate and onlyconnects to WLAN network when the mobile terminal isvery close to the BS of WLAN Secondly at the performancesimulation parts similar to the functional simulation thecoordinates of the car and the BS ofWLAN and LTE networkare the same as the former onesThe biggest difference settingbetween the function and the performance simulation part
Chinese Journal of Engineering 9
0 50 100 150 200025
03
035
04
045
05
055
06
065
07
Simulation time (s)
Load
bal
anci
ng in
dex
TTSProposed methods
(a)
50 100 150 200 250 300 350 400 450 5000
005
01
015
02
025
03
035
04
045
The number of users
Han
doff
call
drop
ping
pro
babi
lity
Proposed methodsTTS
(b)
Figure 11 The performance comparison in different methods
is that it is assumed that another car is moving from points[250 minus50] to points [minus250 minus50] at the same speed with thetarget car
As it is observed in Figure 10(c) at the beginning thecar utilized V2V network to communicate and after then thehandoff process changes the access network from WLAN toLTE according to the rules for choosing the highest valueof APCV (as shown by black lines) among them as accessnetwork The handoff process is consistent with the previoussimulation except that it accessed V2V network initially Bycomparing the throughput of different networks (as shown inFigure 10(d)) it could be found that the network performanceof LTE is the best then there isWLAN and V2V is the worstFrom the aspects of vertical handoff the triangle and trape-zoid shapes (TTS) are always chosen as fuzzy membershipfunctions in many engineering applications Moreover theload balancing degree and handoff call dropping probabilityare used to evaluate the performance of handoff Therefore
after comparison from Figure 11(a) we can find that the loadbalancing degree of our proposed schemes is higher than theTTS In addition as illustrated in Figure 11(b) call droppingprobability is lower in our proposed schemes than that ofTTS especially in the larger number of users conditionsIn general the simulation results validate that our proposedscheme outperforms the existing vertical handoff algorithm
5 Conclusion
This paper proposed a dynamic handoff algorithm in vehiclesnetworks to solve the conflicts between BS and MS such ascall blocking call dropping and ping-pong phenomenonSimulation results validate that the presented joint handoffalgorithm can effectively address the handoff problem amongvarious access networks by taking into account differentscenarios different vehicle speeds and channel conditions
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This research is supported in part by China Impor-tant National Science and Technology Specific Projects(no 2013ZX03001020-002) and by National Key Technol-ogy Research and Development Program of China (no2012BAF14B01) and by National Natural Science Foundationof China (no 61171105 and no 61322110) and by 863 ProgramProject (no 2015AA01A703) and Doctor Funding Program(no 201300051100013)
References
[1] F Siddiqui and S Zeadally ldquoMobility management acrosshybrid wireless networks trends and challengesrdquo IEEE Com-puter Communications vol 11 no 5 pp 45ndash54 2005
[2] N Nasser A Hasswa and H Hassanein ldquoHandoffs in fourthgeneration heterogeneous networksrdquo IEEE CommunicationsMagazine vol 44 no 10 pp 96ndash103 2006
[3] M Ylianttila J Makeela and PMahonen ldquoVehicular telematicsover heterogeneous wireless networks a surveyrdquo ComputerCommunications vol 33 no 7 pp 775ndash793 2010
[4] X Liu L-G Jiang andCHe ldquoVertical handoff algorithmbasedon fuzzy logic in aid of pre-decision methodrdquo Acta ElectronicaSinica vol 35 no 10 pp 1989ndash1993 2007
[5] Y B Kang K Xu Y L Shen et al ldquoOptimal distributed verticalhandoff strategies in vehicular heterogeneous networksrdquo IEEEJournal on Communications supplement 1 pp 69ndash73 2009
[6] D Ma and M Ma ldquoA QoS-based vertical handoff scheme forinterworking of WLAN and WiMAXrdquo in Proceedings of theIEEEGlobal Telecommunications Conference (GLOBECOM rsquo09)pp 1ndash6 Honolulu Hawaii USA December 2009
[7] M Liu Z-C Li X-B Guo and K Zheng ldquoA movementtrend based self-adaptive vertical handoff algorithm and itsperformance evaluationrdquo Chinese Journal of Computers vol 31no 1 pp 112ndash119 2008
[8] Y Li M Li B Cao Y Wang and W Liu ldquoDynamic optimiza-tion of handover parameters adjustment for conflict avoidance
10 Chinese Journal of Engineering
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
networks with various services and coverage ranges So asto accurately obtain the ldquoalways best connectrdquo performancenetworks the joint vertical and horizontal handoff in het-erogeneous networks will become an increasingly importantresearch hotspot although little researche has been done onit In order to allow users to roam among various systemstransparently and seamlessly this paper presents a novelalgorithm which not only supports the horizontal handoffin the same network but also can classify the RSS andoptimize velocity speed and data rate under different channelconditions to choose an optimal network to handoff Forthe sake of low computational complexity RSS is adoptedto handle horizontal handoff and the fuzzy logic decisionalgorithm is used to address vertical handoff There is nodoubt that the fuzzy logic decision algorithm takes many fea-tures of the mobile terminals (MTs) and access network intoconsideration and thus makes the handoff algorithm easierfor hardware implementation The remainder of this paperis organized as follows Section 2 describes the backgroundof handoff process Section 3 indicates the proposed handoffalgorithm and gives the related channel model Numericalresults are presented in Section 4 Finally Section 5 concludesthe whole paper
2 Background
21 The Brief of Handoff It is not difficult to find that the5G wireless access network will consist of access networkswith various coverage and data rates The hybrid networksystem can take advantage of each system to provide highQoS to users According to the relationship between thesource and target network the handoff process can be dividedinto horizontal handoff and vertical handoff as shown inFigure 1 In general horizontal handoff refers to handoffin the same access technology while vertical handoff is thehandoff process occurring between different access networktechnologies No matter which handoff process happens theultimate objective is to keep ldquoalways best connectrdquo (ABC) forusers [11ndash13] That is to say when a mobile user connectsto hybrid access networks concurrently all they will dois to choose a best adjacent cell in the same network oraccess network for its service requirement purpose Inmobilecommunication systems there exists a certain overlap areabetween adjacent cells called handoff belt When the mobileterminal moves into this area the wireless link handoffprocessmust be conducted effectively However in the hybridnetwork systems the coverage area of several networks maybe overlapped Thus a new handoff scenario in the hybridnetwork system has to occur in the near future [14] In orderto minimize the link dropping probability the horizontal andvertical handoff should be considered jointly
22 The Network Architecture and Protocol The existingnetwork architecture consists of mobile terminal accessnetwork and core network In the access network thebase station (eNodeB) bears the control function of wirelesssystem and makes the data transmission rate become fasterMoreover the boundaries between access network and corenetwork are blurredThemobile terminal is connected to the
Horizontalhandoff
LTE
Horizontalhandoff
Horizontalhandoff
WLAN
WAVE
Verticalhandoff
Verticalhandoff
Figure 1 The schematic map of handoff
LTE3G 2G
Corenetwork
Accessnetwork
S1 S1
X2eNodeB
eNodeB
Figure 2 The vehicular wireless communication system
access network via air interfaceThe eNodeB uses S1 interfaceto connect to the core network and X2 interface to link withanother eNodeB respectively as shown in Figure 2
The typical vehicular communication system is dividedinto three subsystems core network vehicle and groundparts The vehicular subsystem mainly includes the vehicu-lar information platform the communication module thesystem multiplexing module the 2G3G automotive microbase station unit and theWi-Fi hotspots equipment It alwaysdirectly communicates with the ground stations throughstandard onboard processing unit which is responsible forthe large capacity data transmission The ground subsystemmainly includes the eNodeB radio remote unit (RRU) andthe S1 and X2 interface optic fibers This paper assumesthat the parts of the core network are related to the accessnetwork Specifically the core network uploads the users
Chinese Journal of Engineering 3
service information and the whole system control informa-tion Moreover the access network is utilized for convertingthe format of 2G3G and separating them into the differentservice networks
Therefore as discussed above there are two types ofhandoff process one is based on X2 interface switch andthe other is connected with S1 interface For the horizontalhandoff the data will be passed throughX2 interface betweeneNodeBs When the user equipment (UE) needs verticalhandoff the source eNodeB will trigger the S1 interface toswitch information from eNodeB to access network and corenetwork
23 The Procedure of Handoff The handoff process has threesteps (1) Measurement of handoff at first UE always triesto find out whether there is a new available network ornot In the vertical handoff process many conditions ofswitch are considered such as signal receiving intensity userpreference network cost and load balance The commonhandoff decision will be triggered when network congestionis serious or current network signal and QoS are weak Thehandoff may also be held when the current network cannotprovide satisfactory businesses to the user (2) Switchingdecision stage there are three kinds of control modes tomake the handoffdecision network control terminal controland terminal auxiliary switch The system will syntheticallytake into account the situation of each alternative networkthe terminal characteristics and the userrsquos preferences forthe current business In some special circumstances otherfactors (such as QoS and system performance security) willalso be considered (3) The execution phase of handoff thisstage is completing the handoff process and switching thecommunication services from the current access point to thetarget network To ensure the completion of the handoffsome protocols are neededmore details about it can be foundin [15]
3 The Handoff Algorithm
In this section we present a novel scheme based on thejoint RSS and fuzzy logic handoff algorithm Figure 3 showsthe algorithm flow chart Assume that the WLAN systemhas successfully established the connection The vehiclemeasurement equipment always detects the various accessnetwork signals Once the RSS is higher than that from otheraccess networks the station keeps the data connection withWLAN and meanwhile rises the horizontal handoff modeOtherwise the mobile station waits for the optimal networkselection time before comparing the RSS again After the RSScomparison if the RSS in WLAN is still higher than theothers the whole handoff process returns to the initial phaseand the vehicle measurement equipment retests the variousaccess network signals Otherwise the algorithmmechanismwill decide to start the vertical handoff mode which makesdecision based on fuzzy inference engine by utilizing Zadehrules and the details are described as follows
31 The Algorithm of Horizontal Handoff During eachperiod the UE in either idle or connection state measures
Select the target cell with the same networkaccording to the processof UE moving
Dynamically adjust the sampling periods duringmeasurement based on the received signal
Horizontal handoffswitching decision
Start the verticalhandoff mode
Measure the relatedparameters ofvarious networkaround the UE
Fuzzy inferenceengine based onZadehrsquos method
Vertical handoffswitching decision
YesNo
Yes
No
Initial phase(connect with the source access network)
Real‐time detection of all kinds of signalsstrength from various networks
Waiting a dwelltime Then ifSsource lt Sselection
handoff modeStart the horizontal
Ssource lt SselectionIf
Figure 3 The flow chart of handoff algorithm
RSSI
Fuzzy inferenceengine based onZadehrsquos method
Data rate Knowledgerules
Reversalfuzzifier
APCVoutput
Interferenceratio
Figure 4 The fuzzy logic decision scheme
the related power parameters related to both the source celland the target one The measurement parameters includereference signal received power (RSRP) reference signalreceived quality (RSRQ) and reference signal strength indi-cator (RSSI) The RSRP is the received power value of thepilot signal from base station It is calculated by the difference
4 Chinese Journal of Engineering
between the reference signal which transmits from the basestation and the path loss
RSRP = 119875119879119909minus PL (dB) (1)
where 119875119879119909
is the transmitter power of the pilot signal and PLrepresents the path loss The RSRQ can be calculated as
RSRQ = 119873 lowast RSRPRSSI
(2)
where 119873 is the number of Resource Blocks (RB) in termsof bandwidth The RSSI is the measurement value of thereceived carrierrsquos power in thewhole systembandwidth how-ever the carrier sends not only data and control informationbut also some interference and noise information In orderto make the handoff decision reasonable the acquired infor-mation from the measurement data needs to be submittedto the Radio Resource Control (RRC) layer In additionthe measured reference signal should be smoothed by filtersbefore being transmitted to the RRC for eliminating randomfluctuations [16] The trigger condition of handoff executionis as follows the QoS of target cell is better than the currentlyserving cell and the difference exceeds a specified thresholdwhen the duration is greater than the trigger delay time Thejudgment criteria are shown as
where 119875target and 119875source represent the strength power of thetarget cell and source cell respectively and 119874
119891means the
offset between related source cell and target cell When themeasurement value of the received signal is consistent with(3) the timer will start working If the relationship describedby (3) continuity satisfies the rule of time to trigger (TTT) thesystemwill determine to drive the handoff process executionThe decision threshold 119867 and TTT have great influenceon the performance of the algorithm Larger 119867 and longerdelay time of departure will make the handoff more difficultTherefore as a result of switch it is easy to lead to linkconnection fail namely drop link If 119867 is too small it willcause frequent switching namely ping-pong phenomenon
32 The Algorithm of Vertical Handoff In this part we willfocus on how to use the algorithm for a vertical operationbased on fuzzy logic As we all know the general elementsof collection set for the membership can only take 0 and 1 In1965 L A Zadeh expanded the membership from only twovalues (0 and 1) to any values within 0 and 1 which was asign that the membership function with fuzzy sets could berepresented by fuzzy probability The fuzzy subset 119861 whichbelongs to the theory domain 119879 is a function characterizedby a collection of membership and can be mapped as follows
120583119861 119879 997888rarr [0 1] (4)
where120583119861means themembership function of the fuzzy subset
120583119861(119905) represents the extent to which the elements 119905 in the
theory domain119879 belong to fuzzy subset 119861The larger value ofit the higher probability of belonging to 119861 For a given theory
TargetBSSource BS
y-axis
x-axis(0 0)
(0 50)(2000 50)
(2000 0)
Figure 5 The horizontal handoff simulation model
LTE-eNodeBWLAN-BS
y-axis
x-axis(0 0) (500 0)
(0 minus50) (250 minus50)
(2200 0)
Figure 6 The vertical handoff simulation model
domain of 119864 a kind of word which is related to 119864 constitutesa set 119875 Its semantics is represented by the function 119877 whichmaps the relationship between set 119875 and 119864 Therefore theword set is a fuzzy function and it can be described as
where 119890 is the element of 119864 And the membership function120583119861(119886 119890) means the extended relationship between 119886 (which
belongs to the set 119875) and 119890 (which belongs to the theorydomain 119864)
120583119877 119875 times 119864 997888rarr [0 1] (6)
In order to deduce the fuzzy logic relationship we willintroduce the principle of Zadehrsquos method if the fuzzyrelationship ldquoIf 119860 then 119861rdquo can be represented by ldquo119860 rarr 119861rdquowhere119860 isin 119880 119861 isin 119881 then the fuzzy logic relationship119877(119906 V)is defined as
Zadeh119877 (119906 V) = (119860 (119906) and 119861 (V)) or (1 minus 119860 (119906)) (7)
where ldquoandrdquo and ldquoorrdquo stand for the supremum and infimumoperatorwith logic relationship respectively For a given119860lowast isin119880 if the relationship matrix 119877 is known then 119861lowast isin 119881 can becalculated by
119861
lowast= 119860
lowast∘ 119877 (8)
where ldquo∘rdquo means the supremum operator with logic relation-ship
The proposed fuzzy logic decision strategy consists ofthree steps (as shown in Figure 4) fuzzy inference enginebased on Zadehrsquos method reversal fuzzifier and APCVoutput At first the input parameters such as RSSI IR andDR are mapped into inference engine by utilizing knowledge
Chinese Journal of Engineering 5
0 500 1000 1500 2000 2500 30000
200
400
600
800
1000
1200
1400
1600
1800
2000
The radius of cell (m)
Aver
age d
wel
l tim
e (s)
5kmh30kmh60kmh
75kmh90kmh120 kmh
(a)
0 2000 4000 6000 8000 100000
01
02
03
04
05
06
07
08
09
1
The radius of cell (m)
Prob
abili
ty o
f han
doff
5kmh30kmh60kmh
75kmh90kmh120 kmh
(b)
Figure 7 The handoff probability and the average time with various cell radius and user speeds
rules And then the inference engine will make a fuzzydecision according to the mapping values But as mentionedabove the output of the fuzzy inference module is still afuzzy set and has no effects on the controlled object directlyTherefore a ldquotranslationrdquo process is needed to transfer a fuzzyvalue to a precise one119891
0This step is called ldquoreversal fuzzifierrdquo
Through comparison with this precise value we can get theaccess point candidacy value (APCV) output The goal ofthis ldquoreversal fuzzifierrdquo is to find the maximum membershipdegree function The ldquoreversal fuzzifierrdquo selects the elementwhich will make the value of membership function thehighest as the output signal And the rules could be describedas
120583 (1198910) ge 120583 (119891) 119891 isin 119865 (9)
In case of nonunique maximum value of the membershipfunction the average value will be taken as follows
1198910=
1
119873
119873
sum
119894=1
119891119894 120583 (119891
119894) ge 120583 (119891) (10)
In different networks different fuzzy inference modelsare adopted The membership function is designed to ensurecomparability among different networks at the same timeAt last in order to complete the behavior of handoff thesystem will select the highest APCV network to accessThereare three parameters that need to be detected in the fuzzymodule RSSI data rate and interference ratio Gaussian and119878-scheme distribution are chosen as the fuzzy membershipbecause of their great performance with respect to real-timecontrol
Figure 9 shows the WLAN membership functions Forexample RSSI stands for the received signal strength indi-cator Therefore we make such an assumption to describe its
knowledge rules (as shown in Figure 9(c)) (i) the member-ship function is 119878-scheme distributionwhenRSSI is in a giveninterval from minus95 dBm to minus80 dBm (ii) if RSSI belongs to theinterval [minus90 dBmminus70 dBm] then themembership functionfollows Gaussian distribution and (iii) once the value ofRSSI is higher than minus80 dBm the membership function is119878-scheme distribution as well The fuzzy set of RSSI the setof data rate and the fuzzy set of interference ratio in WLANare [good normal weak] with value set [3 2 1] Hence theoutput fuzzy sets are mapped as
119878 =
3
sum
119894=1
119872119894119896 (11)
Take a regular WLAN network as an example if RSSI isweak and data rate is weak and interference ratio is weakthen 119878 is 3 For the WLAN network each input has threefuzzy sequences so there are 27 fuzzy rules in the knowledgerules Due to the sensitivity to UE speed in V2V network5 fuzzy variable sequences are used (45 fuzzy rules) in theknowledge rules The input parameters in the fuzzy logicmodule are mapped into different fuzzy sets and the outputwould be acquired utilizing the maximum membershipdegree function method The fuzzy logic outputs decisionvalue indicates that the stable degree of difference betweentwo network decisions is lower than the threshold level thusthe connection with the source network cell will be held onOtherwise the handoff executionwill be done if the hysteresistime is longer than a certain value that is
VALUE TARGET minus VALUE SOURCE ge Hysteresis
VALUE SOURCE le Threshold(12)
6 Chinese Journal of Engineering
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The s
treng
th o
f rec
eive
d sig
nal
from
targ
et ce
ll (d
Bm)
The s
treng
th o
f rec
eive
d sig
nal
from
sour
ce ce
ll (d
Bm)
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The location of vehicle (km)
The location of vehicle (km)
(a) The strength of received signal from source and target cell
06 07 08 09 1 11 12 13 140
01
02
03
04
05
06
07
08
09
1
The location of vehicle (km)
The p
roba
bilit
y of
han
doff
The theoretical handoff point
The probability of connectingwith the target cell
The probability of connectingwith the source cell
(b) The probability of handoff
06 07 08 09 1 11 12 13 1401
02
03
04
05
06
07
08
09
10
The location of vehicle (km)
Prob
abili
ty o
f han
doff The actual handoff location point
(c) The actual handoff location point
20 40 60 80 100 120 140 160 1800
005
01
015
02
025
03
035
04
The speed of mobile terminal (kmh)
Ping
-pon
g ha
ndoff
ratio
Proposed algorithmExisting RSSI algorithm(d) The performance comparison
Figure 8 The horizontal handoff simulation results
33 Channel Model This subsection will adopt the COST 231Hata channel model to realize the real-time simulation [1718] which is depicted as
PL (dB) = (397 minus 701 log10 (ℎBS)) log10 (119889
1000
)
+ 429 + (3654 minus 12ℎMS) log10 (119891119888)
minus 1298 log10(ℎBS) + 074ℎMS + 119862
(13)
where ℎBS and ℎMS represent the height of BS and MSrespectively 119889 means the distance between BS and MS 119891119888denotes the center frequency and 119862 is a constant
The channel impulse response (CIR) from transmitterantenna element 119904 to receiver element 119906 for cluster 119899 isexpressed as
is the cross polarization power ratio in linear scale V is themoving speed of user and 120579
119899119898AoA and 120579119899119898AoD mean the
angle of arrival (AoA) and angle of departure (AoD) with the119898th subpath in 119899th path respectively Assume that there are6 paths and each path includes 20 subpaths
4 The Simulation and Validation
To evaluate the realistic performance of our proposed algo-rithm the simulation model scenarios and various parame-ters are developed using MATLAB The performance of hor-izontal and vertical handoff algorithms is tested separately
Figure 7 compares the handoff probability and averagedwell time with different moving speed and cellular radius Itis found that the expansion of cellular radius and decreasingvehicle moving speed will lead to increase of average dwelltime and channel holding time In particular the relationshipbetween the coverage radius and dwell time is obviouslylinear in the low vehicle speed regions On the contrary withthe increase of vehicle speed the handoff probability willcontinue to decrease
The horizontal handoff is simulated using the followingparameters (as shown in Figure 5) the point coordinates of
source BS and target BS are [0 50] and [2000 50] whichmeans that the distance between the source BS and targetBS is 2000m and the distance between road and BS is50m the vehicle speed is 80 kmh and is moving from [0 0]to [2000 0] the heights of BS and MS are 30m and 1mseparately the transmitted power is 44 dBm and the standarddeviation of channel shadow fading is 8 dB
Figure 8 shows the simulation results of horizontal hand-off It is clearly seen that when the vehiclemoves from [500 0]to [1500 0] the strength of received signal from source BSis reducing while the signal level from target BS is growingMeanwhile the handoff probability shows a similar trendand the handoff should occur at the points near [1000 0]according to the theoretical judgment of holding probabilityIt can also be observed that the handoff happens at the peakof the curve namely the coordinate of the vehicle is [1055 0]There is a little far distance from the actual handoff locationto the theoretical one due to the hysteresis effect of dwelltime on the joint handoff algorithm As we all know theexisting RSSI algorithm depends on the decisions of priorityon the mobile node In particular the priority is dividedinto low priority and high priority The low priority whichis below certain level estimate receives signal strength from
8 Chinese Journal of Engineering
0 50 100 150 200 250 300 350 400minus105
minus100
minus95
minus90
minus85
minus80
minus75
minus70
The simulation time (s)
RSS
(dBm
)The received signal strength
LTEWLAN
(a) RSS
0 50 100 150 200 250 3000
01
02
03
04
05
06
07
08
09
1
The simulation time (s)
Mar
k of
net
wor
k
LTE
WLANThe state of handoff
(b) The state of handoff
50 100 150 200 250 300 3500
1
2
3
4
5
6
7
8
9
10
The simulation time (s)
APC
V
LTEWLANV2V
V2VWLAN
LTE
(c) APCV and state of handoff
1 15 2 25 3 35 402
03
04
05
06
07
08
09
1
Request (MBs)
Thro
ughp
ut
V2VWLANLTE
(d) Throughput versus request
Figure 10 The vertical handoff simulation results
another cell before it starts handoff in the cell ThereforeFigure 8(d) presents a performance comparison between theexisting RSSI horizontal handoff algorithm and our proposedschemes It is clear that no matter how much the speedof users is the ping-pong handoff ratio of existing RSSIhorizontal handoff algorithm is higher than the proposedschemes It proves that ping-pong effect can be avoidedeffectively by using dwell timer and hysteresis parameters
The vertical handoff simulation is divided into two partsone focuses on testing whether the handoff could be appliedautomatically and another part analyzes the performanceof the proposed algorithm Firstly assume a car is movingfrom the initial points [0 minus50] to [2500 minus50] at the speed of45 kmh and the coordinates of WLAN-BS and LTE-eNodeare [500 0] and [2200 0] respectively (as shown in Figure 6)
Figure 10(a) shows the alternative trend of the RSS during thewhole movement and Figure 10(b) shows the actual handoffresults It is not difficult to find such a handoff process thecar accessed LTE at first and then changed the network toWLAN because it ran close to theWLAN-BS when it left thecoverage of WLAN the handoff happened again to replaceWLAN by LTE signals Obviously because of different pathloss and shadow fading due to different center frequencythe vehicle utilizes LTE network to communicate and onlyconnects to WLAN network when the mobile terminal isvery close to the BS of WLAN Secondly at the performancesimulation parts similar to the functional simulation thecoordinates of the car and the BS ofWLAN and LTE networkare the same as the former onesThe biggest difference settingbetween the function and the performance simulation part
Chinese Journal of Engineering 9
0 50 100 150 200025
03
035
04
045
05
055
06
065
07
Simulation time (s)
Load
bal
anci
ng in
dex
TTSProposed methods
(a)
50 100 150 200 250 300 350 400 450 5000
005
01
015
02
025
03
035
04
045
The number of users
Han
doff
call
drop
ping
pro
babi
lity
Proposed methodsTTS
(b)
Figure 11 The performance comparison in different methods
is that it is assumed that another car is moving from points[250 minus50] to points [minus250 minus50] at the same speed with thetarget car
As it is observed in Figure 10(c) at the beginning thecar utilized V2V network to communicate and after then thehandoff process changes the access network from WLAN toLTE according to the rules for choosing the highest valueof APCV (as shown by black lines) among them as accessnetwork The handoff process is consistent with the previoussimulation except that it accessed V2V network initially Bycomparing the throughput of different networks (as shown inFigure 10(d)) it could be found that the network performanceof LTE is the best then there isWLAN and V2V is the worstFrom the aspects of vertical handoff the triangle and trape-zoid shapes (TTS) are always chosen as fuzzy membershipfunctions in many engineering applications Moreover theload balancing degree and handoff call dropping probabilityare used to evaluate the performance of handoff Therefore
after comparison from Figure 11(a) we can find that the loadbalancing degree of our proposed schemes is higher than theTTS In addition as illustrated in Figure 11(b) call droppingprobability is lower in our proposed schemes than that ofTTS especially in the larger number of users conditionsIn general the simulation results validate that our proposedscheme outperforms the existing vertical handoff algorithm
5 Conclusion
This paper proposed a dynamic handoff algorithm in vehiclesnetworks to solve the conflicts between BS and MS such ascall blocking call dropping and ping-pong phenomenonSimulation results validate that the presented joint handoffalgorithm can effectively address the handoff problem amongvarious access networks by taking into account differentscenarios different vehicle speeds and channel conditions
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This research is supported in part by China Impor-tant National Science and Technology Specific Projects(no 2013ZX03001020-002) and by National Key Technol-ogy Research and Development Program of China (no2012BAF14B01) and by National Natural Science Foundationof China (no 61171105 and no 61322110) and by 863 ProgramProject (no 2015AA01A703) and Doctor Funding Program(no 201300051100013)
References
[1] F Siddiqui and S Zeadally ldquoMobility management acrosshybrid wireless networks trends and challengesrdquo IEEE Com-puter Communications vol 11 no 5 pp 45ndash54 2005
[2] N Nasser A Hasswa and H Hassanein ldquoHandoffs in fourthgeneration heterogeneous networksrdquo IEEE CommunicationsMagazine vol 44 no 10 pp 96ndash103 2006
[3] M Ylianttila J Makeela and PMahonen ldquoVehicular telematicsover heterogeneous wireless networks a surveyrdquo ComputerCommunications vol 33 no 7 pp 775ndash793 2010
[4] X Liu L-G Jiang andCHe ldquoVertical handoff algorithmbasedon fuzzy logic in aid of pre-decision methodrdquo Acta ElectronicaSinica vol 35 no 10 pp 1989ndash1993 2007
[5] Y B Kang K Xu Y L Shen et al ldquoOptimal distributed verticalhandoff strategies in vehicular heterogeneous networksrdquo IEEEJournal on Communications supplement 1 pp 69ndash73 2009
[6] D Ma and M Ma ldquoA QoS-based vertical handoff scheme forinterworking of WLAN and WiMAXrdquo in Proceedings of theIEEEGlobal Telecommunications Conference (GLOBECOM rsquo09)pp 1ndash6 Honolulu Hawaii USA December 2009
[7] M Liu Z-C Li X-B Guo and K Zheng ldquoA movementtrend based self-adaptive vertical handoff algorithm and itsperformance evaluationrdquo Chinese Journal of Computers vol 31no 1 pp 112ndash119 2008
[8] Y Li M Li B Cao Y Wang and W Liu ldquoDynamic optimiza-tion of handover parameters adjustment for conflict avoidance
10 Chinese Journal of Engineering
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
service information and the whole system control informa-tion Moreover the access network is utilized for convertingthe format of 2G3G and separating them into the differentservice networks
Therefore as discussed above there are two types ofhandoff process one is based on X2 interface switch andthe other is connected with S1 interface For the horizontalhandoff the data will be passed throughX2 interface betweeneNodeBs When the user equipment (UE) needs verticalhandoff the source eNodeB will trigger the S1 interface toswitch information from eNodeB to access network and corenetwork
23 The Procedure of Handoff The handoff process has threesteps (1) Measurement of handoff at first UE always triesto find out whether there is a new available network ornot In the vertical handoff process many conditions ofswitch are considered such as signal receiving intensity userpreference network cost and load balance The commonhandoff decision will be triggered when network congestionis serious or current network signal and QoS are weak Thehandoff may also be held when the current network cannotprovide satisfactory businesses to the user (2) Switchingdecision stage there are three kinds of control modes tomake the handoffdecision network control terminal controland terminal auxiliary switch The system will syntheticallytake into account the situation of each alternative networkthe terminal characteristics and the userrsquos preferences forthe current business In some special circumstances otherfactors (such as QoS and system performance security) willalso be considered (3) The execution phase of handoff thisstage is completing the handoff process and switching thecommunication services from the current access point to thetarget network To ensure the completion of the handoffsome protocols are neededmore details about it can be foundin [15]
3 The Handoff Algorithm
In this section we present a novel scheme based on thejoint RSS and fuzzy logic handoff algorithm Figure 3 showsthe algorithm flow chart Assume that the WLAN systemhas successfully established the connection The vehiclemeasurement equipment always detects the various accessnetwork signals Once the RSS is higher than that from otheraccess networks the station keeps the data connection withWLAN and meanwhile rises the horizontal handoff modeOtherwise the mobile station waits for the optimal networkselection time before comparing the RSS again After the RSScomparison if the RSS in WLAN is still higher than theothers the whole handoff process returns to the initial phaseand the vehicle measurement equipment retests the variousaccess network signals Otherwise the algorithmmechanismwill decide to start the vertical handoff mode which makesdecision based on fuzzy inference engine by utilizing Zadehrules and the details are described as follows
31 The Algorithm of Horizontal Handoff During eachperiod the UE in either idle or connection state measures
Select the target cell with the same networkaccording to the processof UE moving
Dynamically adjust the sampling periods duringmeasurement based on the received signal
Horizontal handoffswitching decision
Start the verticalhandoff mode
Measure the relatedparameters ofvarious networkaround the UE
Fuzzy inferenceengine based onZadehrsquos method
Vertical handoffswitching decision
YesNo
Yes
No
Initial phase(connect with the source access network)
Real‐time detection of all kinds of signalsstrength from various networks
Waiting a dwelltime Then ifSsource lt Sselection
handoff modeStart the horizontal
Ssource lt SselectionIf
Figure 3 The flow chart of handoff algorithm
RSSI
Fuzzy inferenceengine based onZadehrsquos method
Data rate Knowledgerules
Reversalfuzzifier
APCVoutput
Interferenceratio
Figure 4 The fuzzy logic decision scheme
the related power parameters related to both the source celland the target one The measurement parameters includereference signal received power (RSRP) reference signalreceived quality (RSRQ) and reference signal strength indi-cator (RSSI) The RSRP is the received power value of thepilot signal from base station It is calculated by the difference
4 Chinese Journal of Engineering
between the reference signal which transmits from the basestation and the path loss
RSRP = 119875119879119909minus PL (dB) (1)
where 119875119879119909
is the transmitter power of the pilot signal and PLrepresents the path loss The RSRQ can be calculated as
RSRQ = 119873 lowast RSRPRSSI
(2)
where 119873 is the number of Resource Blocks (RB) in termsof bandwidth The RSSI is the measurement value of thereceived carrierrsquos power in thewhole systembandwidth how-ever the carrier sends not only data and control informationbut also some interference and noise information In orderto make the handoff decision reasonable the acquired infor-mation from the measurement data needs to be submittedto the Radio Resource Control (RRC) layer In additionthe measured reference signal should be smoothed by filtersbefore being transmitted to the RRC for eliminating randomfluctuations [16] The trigger condition of handoff executionis as follows the QoS of target cell is better than the currentlyserving cell and the difference exceeds a specified thresholdwhen the duration is greater than the trigger delay time Thejudgment criteria are shown as
where 119875target and 119875source represent the strength power of thetarget cell and source cell respectively and 119874
119891means the
offset between related source cell and target cell When themeasurement value of the received signal is consistent with(3) the timer will start working If the relationship describedby (3) continuity satisfies the rule of time to trigger (TTT) thesystemwill determine to drive the handoff process executionThe decision threshold 119867 and TTT have great influenceon the performance of the algorithm Larger 119867 and longerdelay time of departure will make the handoff more difficultTherefore as a result of switch it is easy to lead to linkconnection fail namely drop link If 119867 is too small it willcause frequent switching namely ping-pong phenomenon
32 The Algorithm of Vertical Handoff In this part we willfocus on how to use the algorithm for a vertical operationbased on fuzzy logic As we all know the general elementsof collection set for the membership can only take 0 and 1 In1965 L A Zadeh expanded the membership from only twovalues (0 and 1) to any values within 0 and 1 which was asign that the membership function with fuzzy sets could berepresented by fuzzy probability The fuzzy subset 119861 whichbelongs to the theory domain 119879 is a function characterizedby a collection of membership and can be mapped as follows
120583119861 119879 997888rarr [0 1] (4)
where120583119861means themembership function of the fuzzy subset
120583119861(119905) represents the extent to which the elements 119905 in the
theory domain119879 belong to fuzzy subset 119861The larger value ofit the higher probability of belonging to 119861 For a given theory
TargetBSSource BS
y-axis
x-axis(0 0)
(0 50)(2000 50)
(2000 0)
Figure 5 The horizontal handoff simulation model
LTE-eNodeBWLAN-BS
y-axis
x-axis(0 0) (500 0)
(0 minus50) (250 minus50)
(2200 0)
Figure 6 The vertical handoff simulation model
domain of 119864 a kind of word which is related to 119864 constitutesa set 119875 Its semantics is represented by the function 119877 whichmaps the relationship between set 119875 and 119864 Therefore theword set is a fuzzy function and it can be described as
where 119890 is the element of 119864 And the membership function120583119861(119886 119890) means the extended relationship between 119886 (which
belongs to the set 119875) and 119890 (which belongs to the theorydomain 119864)
120583119877 119875 times 119864 997888rarr [0 1] (6)
In order to deduce the fuzzy logic relationship we willintroduce the principle of Zadehrsquos method if the fuzzyrelationship ldquoIf 119860 then 119861rdquo can be represented by ldquo119860 rarr 119861rdquowhere119860 isin 119880 119861 isin 119881 then the fuzzy logic relationship119877(119906 V)is defined as
Zadeh119877 (119906 V) = (119860 (119906) and 119861 (V)) or (1 minus 119860 (119906)) (7)
where ldquoandrdquo and ldquoorrdquo stand for the supremum and infimumoperatorwith logic relationship respectively For a given119860lowast isin119880 if the relationship matrix 119877 is known then 119861lowast isin 119881 can becalculated by
119861
lowast= 119860
lowast∘ 119877 (8)
where ldquo∘rdquo means the supremum operator with logic relation-ship
The proposed fuzzy logic decision strategy consists ofthree steps (as shown in Figure 4) fuzzy inference enginebased on Zadehrsquos method reversal fuzzifier and APCVoutput At first the input parameters such as RSSI IR andDR are mapped into inference engine by utilizing knowledge
Chinese Journal of Engineering 5
0 500 1000 1500 2000 2500 30000
200
400
600
800
1000
1200
1400
1600
1800
2000
The radius of cell (m)
Aver
age d
wel
l tim
e (s)
5kmh30kmh60kmh
75kmh90kmh120 kmh
(a)
0 2000 4000 6000 8000 100000
01
02
03
04
05
06
07
08
09
1
The radius of cell (m)
Prob
abili
ty o
f han
doff
5kmh30kmh60kmh
75kmh90kmh120 kmh
(b)
Figure 7 The handoff probability and the average time with various cell radius and user speeds
rules And then the inference engine will make a fuzzydecision according to the mapping values But as mentionedabove the output of the fuzzy inference module is still afuzzy set and has no effects on the controlled object directlyTherefore a ldquotranslationrdquo process is needed to transfer a fuzzyvalue to a precise one119891
0This step is called ldquoreversal fuzzifierrdquo
Through comparison with this precise value we can get theaccess point candidacy value (APCV) output The goal ofthis ldquoreversal fuzzifierrdquo is to find the maximum membershipdegree function The ldquoreversal fuzzifierrdquo selects the elementwhich will make the value of membership function thehighest as the output signal And the rules could be describedas
120583 (1198910) ge 120583 (119891) 119891 isin 119865 (9)
In case of nonunique maximum value of the membershipfunction the average value will be taken as follows
1198910=
1
119873
119873
sum
119894=1
119891119894 120583 (119891
119894) ge 120583 (119891) (10)
In different networks different fuzzy inference modelsare adopted The membership function is designed to ensurecomparability among different networks at the same timeAt last in order to complete the behavior of handoff thesystem will select the highest APCV network to accessThereare three parameters that need to be detected in the fuzzymodule RSSI data rate and interference ratio Gaussian and119878-scheme distribution are chosen as the fuzzy membershipbecause of their great performance with respect to real-timecontrol
Figure 9 shows the WLAN membership functions Forexample RSSI stands for the received signal strength indi-cator Therefore we make such an assumption to describe its
knowledge rules (as shown in Figure 9(c)) (i) the member-ship function is 119878-scheme distributionwhenRSSI is in a giveninterval from minus95 dBm to minus80 dBm (ii) if RSSI belongs to theinterval [minus90 dBmminus70 dBm] then themembership functionfollows Gaussian distribution and (iii) once the value ofRSSI is higher than minus80 dBm the membership function is119878-scheme distribution as well The fuzzy set of RSSI the setof data rate and the fuzzy set of interference ratio in WLANare [good normal weak] with value set [3 2 1] Hence theoutput fuzzy sets are mapped as
119878 =
3
sum
119894=1
119872119894119896 (11)
Take a regular WLAN network as an example if RSSI isweak and data rate is weak and interference ratio is weakthen 119878 is 3 For the WLAN network each input has threefuzzy sequences so there are 27 fuzzy rules in the knowledgerules Due to the sensitivity to UE speed in V2V network5 fuzzy variable sequences are used (45 fuzzy rules) in theknowledge rules The input parameters in the fuzzy logicmodule are mapped into different fuzzy sets and the outputwould be acquired utilizing the maximum membershipdegree function method The fuzzy logic outputs decisionvalue indicates that the stable degree of difference betweentwo network decisions is lower than the threshold level thusthe connection with the source network cell will be held onOtherwise the handoff executionwill be done if the hysteresistime is longer than a certain value that is
VALUE TARGET minus VALUE SOURCE ge Hysteresis
VALUE SOURCE le Threshold(12)
6 Chinese Journal of Engineering
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The s
treng
th o
f rec
eive
d sig
nal
from
targ
et ce
ll (d
Bm)
The s
treng
th o
f rec
eive
d sig
nal
from
sour
ce ce
ll (d
Bm)
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The location of vehicle (km)
The location of vehicle (km)
(a) The strength of received signal from source and target cell
06 07 08 09 1 11 12 13 140
01
02
03
04
05
06
07
08
09
1
The location of vehicle (km)
The p
roba
bilit
y of
han
doff
The theoretical handoff point
The probability of connectingwith the target cell
The probability of connectingwith the source cell
(b) The probability of handoff
06 07 08 09 1 11 12 13 1401
02
03
04
05
06
07
08
09
10
The location of vehicle (km)
Prob
abili
ty o
f han
doff The actual handoff location point
(c) The actual handoff location point
20 40 60 80 100 120 140 160 1800
005
01
015
02
025
03
035
04
The speed of mobile terminal (kmh)
Ping
-pon
g ha
ndoff
ratio
Proposed algorithmExisting RSSI algorithm(d) The performance comparison
Figure 8 The horizontal handoff simulation results
33 Channel Model This subsection will adopt the COST 231Hata channel model to realize the real-time simulation [1718] which is depicted as
PL (dB) = (397 minus 701 log10 (ℎBS)) log10 (119889
1000
)
+ 429 + (3654 minus 12ℎMS) log10 (119891119888)
minus 1298 log10(ℎBS) + 074ℎMS + 119862
(13)
where ℎBS and ℎMS represent the height of BS and MSrespectively 119889 means the distance between BS and MS 119891119888denotes the center frequency and 119862 is a constant
The channel impulse response (CIR) from transmitterantenna element 119904 to receiver element 119906 for cluster 119899 isexpressed as
is the cross polarization power ratio in linear scale V is themoving speed of user and 120579
119899119898AoA and 120579119899119898AoD mean the
angle of arrival (AoA) and angle of departure (AoD) with the119898th subpath in 119899th path respectively Assume that there are6 paths and each path includes 20 subpaths
4 The Simulation and Validation
To evaluate the realistic performance of our proposed algo-rithm the simulation model scenarios and various parame-ters are developed using MATLAB The performance of hor-izontal and vertical handoff algorithms is tested separately
Figure 7 compares the handoff probability and averagedwell time with different moving speed and cellular radius Itis found that the expansion of cellular radius and decreasingvehicle moving speed will lead to increase of average dwelltime and channel holding time In particular the relationshipbetween the coverage radius and dwell time is obviouslylinear in the low vehicle speed regions On the contrary withthe increase of vehicle speed the handoff probability willcontinue to decrease
The horizontal handoff is simulated using the followingparameters (as shown in Figure 5) the point coordinates of
source BS and target BS are [0 50] and [2000 50] whichmeans that the distance between the source BS and targetBS is 2000m and the distance between road and BS is50m the vehicle speed is 80 kmh and is moving from [0 0]to [2000 0] the heights of BS and MS are 30m and 1mseparately the transmitted power is 44 dBm and the standarddeviation of channel shadow fading is 8 dB
Figure 8 shows the simulation results of horizontal hand-off It is clearly seen that when the vehiclemoves from [500 0]to [1500 0] the strength of received signal from source BSis reducing while the signal level from target BS is growingMeanwhile the handoff probability shows a similar trendand the handoff should occur at the points near [1000 0]according to the theoretical judgment of holding probabilityIt can also be observed that the handoff happens at the peakof the curve namely the coordinate of the vehicle is [1055 0]There is a little far distance from the actual handoff locationto the theoretical one due to the hysteresis effect of dwelltime on the joint handoff algorithm As we all know theexisting RSSI algorithm depends on the decisions of priorityon the mobile node In particular the priority is dividedinto low priority and high priority The low priority whichis below certain level estimate receives signal strength from
8 Chinese Journal of Engineering
0 50 100 150 200 250 300 350 400minus105
minus100
minus95
minus90
minus85
minus80
minus75
minus70
The simulation time (s)
RSS
(dBm
)The received signal strength
LTEWLAN
(a) RSS
0 50 100 150 200 250 3000
01
02
03
04
05
06
07
08
09
1
The simulation time (s)
Mar
k of
net
wor
k
LTE
WLANThe state of handoff
(b) The state of handoff
50 100 150 200 250 300 3500
1
2
3
4
5
6
7
8
9
10
The simulation time (s)
APC
V
LTEWLANV2V
V2VWLAN
LTE
(c) APCV and state of handoff
1 15 2 25 3 35 402
03
04
05
06
07
08
09
1
Request (MBs)
Thro
ughp
ut
V2VWLANLTE
(d) Throughput versus request
Figure 10 The vertical handoff simulation results
another cell before it starts handoff in the cell ThereforeFigure 8(d) presents a performance comparison between theexisting RSSI horizontal handoff algorithm and our proposedschemes It is clear that no matter how much the speedof users is the ping-pong handoff ratio of existing RSSIhorizontal handoff algorithm is higher than the proposedschemes It proves that ping-pong effect can be avoidedeffectively by using dwell timer and hysteresis parameters
The vertical handoff simulation is divided into two partsone focuses on testing whether the handoff could be appliedautomatically and another part analyzes the performanceof the proposed algorithm Firstly assume a car is movingfrom the initial points [0 minus50] to [2500 minus50] at the speed of45 kmh and the coordinates of WLAN-BS and LTE-eNodeare [500 0] and [2200 0] respectively (as shown in Figure 6)
Figure 10(a) shows the alternative trend of the RSS during thewhole movement and Figure 10(b) shows the actual handoffresults It is not difficult to find such a handoff process thecar accessed LTE at first and then changed the network toWLAN because it ran close to theWLAN-BS when it left thecoverage of WLAN the handoff happened again to replaceWLAN by LTE signals Obviously because of different pathloss and shadow fading due to different center frequencythe vehicle utilizes LTE network to communicate and onlyconnects to WLAN network when the mobile terminal isvery close to the BS of WLAN Secondly at the performancesimulation parts similar to the functional simulation thecoordinates of the car and the BS ofWLAN and LTE networkare the same as the former onesThe biggest difference settingbetween the function and the performance simulation part
Chinese Journal of Engineering 9
0 50 100 150 200025
03
035
04
045
05
055
06
065
07
Simulation time (s)
Load
bal
anci
ng in
dex
TTSProposed methods
(a)
50 100 150 200 250 300 350 400 450 5000
005
01
015
02
025
03
035
04
045
The number of users
Han
doff
call
drop
ping
pro
babi
lity
Proposed methodsTTS
(b)
Figure 11 The performance comparison in different methods
is that it is assumed that another car is moving from points[250 minus50] to points [minus250 minus50] at the same speed with thetarget car
As it is observed in Figure 10(c) at the beginning thecar utilized V2V network to communicate and after then thehandoff process changes the access network from WLAN toLTE according to the rules for choosing the highest valueof APCV (as shown by black lines) among them as accessnetwork The handoff process is consistent with the previoussimulation except that it accessed V2V network initially Bycomparing the throughput of different networks (as shown inFigure 10(d)) it could be found that the network performanceof LTE is the best then there isWLAN and V2V is the worstFrom the aspects of vertical handoff the triangle and trape-zoid shapes (TTS) are always chosen as fuzzy membershipfunctions in many engineering applications Moreover theload balancing degree and handoff call dropping probabilityare used to evaluate the performance of handoff Therefore
after comparison from Figure 11(a) we can find that the loadbalancing degree of our proposed schemes is higher than theTTS In addition as illustrated in Figure 11(b) call droppingprobability is lower in our proposed schemes than that ofTTS especially in the larger number of users conditionsIn general the simulation results validate that our proposedscheme outperforms the existing vertical handoff algorithm
5 Conclusion
This paper proposed a dynamic handoff algorithm in vehiclesnetworks to solve the conflicts between BS and MS such ascall blocking call dropping and ping-pong phenomenonSimulation results validate that the presented joint handoffalgorithm can effectively address the handoff problem amongvarious access networks by taking into account differentscenarios different vehicle speeds and channel conditions
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This research is supported in part by China Impor-tant National Science and Technology Specific Projects(no 2013ZX03001020-002) and by National Key Technol-ogy Research and Development Program of China (no2012BAF14B01) and by National Natural Science Foundationof China (no 61171105 and no 61322110) and by 863 ProgramProject (no 2015AA01A703) and Doctor Funding Program(no 201300051100013)
References
[1] F Siddiqui and S Zeadally ldquoMobility management acrosshybrid wireless networks trends and challengesrdquo IEEE Com-puter Communications vol 11 no 5 pp 45ndash54 2005
[2] N Nasser A Hasswa and H Hassanein ldquoHandoffs in fourthgeneration heterogeneous networksrdquo IEEE CommunicationsMagazine vol 44 no 10 pp 96ndash103 2006
[3] M Ylianttila J Makeela and PMahonen ldquoVehicular telematicsover heterogeneous wireless networks a surveyrdquo ComputerCommunications vol 33 no 7 pp 775ndash793 2010
[4] X Liu L-G Jiang andCHe ldquoVertical handoff algorithmbasedon fuzzy logic in aid of pre-decision methodrdquo Acta ElectronicaSinica vol 35 no 10 pp 1989ndash1993 2007
[5] Y B Kang K Xu Y L Shen et al ldquoOptimal distributed verticalhandoff strategies in vehicular heterogeneous networksrdquo IEEEJournal on Communications supplement 1 pp 69ndash73 2009
[6] D Ma and M Ma ldquoA QoS-based vertical handoff scheme forinterworking of WLAN and WiMAXrdquo in Proceedings of theIEEEGlobal Telecommunications Conference (GLOBECOM rsquo09)pp 1ndash6 Honolulu Hawaii USA December 2009
[7] M Liu Z-C Li X-B Guo and K Zheng ldquoA movementtrend based self-adaptive vertical handoff algorithm and itsperformance evaluationrdquo Chinese Journal of Computers vol 31no 1 pp 112ndash119 2008
[8] Y Li M Li B Cao Y Wang and W Liu ldquoDynamic optimiza-tion of handover parameters adjustment for conflict avoidance
10 Chinese Journal of Engineering
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
between the reference signal which transmits from the basestation and the path loss
RSRP = 119875119879119909minus PL (dB) (1)
where 119875119879119909
is the transmitter power of the pilot signal and PLrepresents the path loss The RSRQ can be calculated as
RSRQ = 119873 lowast RSRPRSSI
(2)
where 119873 is the number of Resource Blocks (RB) in termsof bandwidth The RSSI is the measurement value of thereceived carrierrsquos power in thewhole systembandwidth how-ever the carrier sends not only data and control informationbut also some interference and noise information In orderto make the handoff decision reasonable the acquired infor-mation from the measurement data needs to be submittedto the Radio Resource Control (RRC) layer In additionthe measured reference signal should be smoothed by filtersbefore being transmitted to the RRC for eliminating randomfluctuations [16] The trigger condition of handoff executionis as follows the QoS of target cell is better than the currentlyserving cell and the difference exceeds a specified thresholdwhen the duration is greater than the trigger delay time Thejudgment criteria are shown as
where 119875target and 119875source represent the strength power of thetarget cell and source cell respectively and 119874
119891means the
offset between related source cell and target cell When themeasurement value of the received signal is consistent with(3) the timer will start working If the relationship describedby (3) continuity satisfies the rule of time to trigger (TTT) thesystemwill determine to drive the handoff process executionThe decision threshold 119867 and TTT have great influenceon the performance of the algorithm Larger 119867 and longerdelay time of departure will make the handoff more difficultTherefore as a result of switch it is easy to lead to linkconnection fail namely drop link If 119867 is too small it willcause frequent switching namely ping-pong phenomenon
32 The Algorithm of Vertical Handoff In this part we willfocus on how to use the algorithm for a vertical operationbased on fuzzy logic As we all know the general elementsof collection set for the membership can only take 0 and 1 In1965 L A Zadeh expanded the membership from only twovalues (0 and 1) to any values within 0 and 1 which was asign that the membership function with fuzzy sets could berepresented by fuzzy probability The fuzzy subset 119861 whichbelongs to the theory domain 119879 is a function characterizedby a collection of membership and can be mapped as follows
120583119861 119879 997888rarr [0 1] (4)
where120583119861means themembership function of the fuzzy subset
120583119861(119905) represents the extent to which the elements 119905 in the
theory domain119879 belong to fuzzy subset 119861The larger value ofit the higher probability of belonging to 119861 For a given theory
TargetBSSource BS
y-axis
x-axis(0 0)
(0 50)(2000 50)
(2000 0)
Figure 5 The horizontal handoff simulation model
LTE-eNodeBWLAN-BS
y-axis
x-axis(0 0) (500 0)
(0 minus50) (250 minus50)
(2200 0)
Figure 6 The vertical handoff simulation model
domain of 119864 a kind of word which is related to 119864 constitutesa set 119875 Its semantics is represented by the function 119877 whichmaps the relationship between set 119875 and 119864 Therefore theword set is a fuzzy function and it can be described as
where 119890 is the element of 119864 And the membership function120583119861(119886 119890) means the extended relationship between 119886 (which
belongs to the set 119875) and 119890 (which belongs to the theorydomain 119864)
120583119877 119875 times 119864 997888rarr [0 1] (6)
In order to deduce the fuzzy logic relationship we willintroduce the principle of Zadehrsquos method if the fuzzyrelationship ldquoIf 119860 then 119861rdquo can be represented by ldquo119860 rarr 119861rdquowhere119860 isin 119880 119861 isin 119881 then the fuzzy logic relationship119877(119906 V)is defined as
Zadeh119877 (119906 V) = (119860 (119906) and 119861 (V)) or (1 minus 119860 (119906)) (7)
where ldquoandrdquo and ldquoorrdquo stand for the supremum and infimumoperatorwith logic relationship respectively For a given119860lowast isin119880 if the relationship matrix 119877 is known then 119861lowast isin 119881 can becalculated by
119861
lowast= 119860
lowast∘ 119877 (8)
where ldquo∘rdquo means the supremum operator with logic relation-ship
The proposed fuzzy logic decision strategy consists ofthree steps (as shown in Figure 4) fuzzy inference enginebased on Zadehrsquos method reversal fuzzifier and APCVoutput At first the input parameters such as RSSI IR andDR are mapped into inference engine by utilizing knowledge
Chinese Journal of Engineering 5
0 500 1000 1500 2000 2500 30000
200
400
600
800
1000
1200
1400
1600
1800
2000
The radius of cell (m)
Aver
age d
wel
l tim
e (s)
5kmh30kmh60kmh
75kmh90kmh120 kmh
(a)
0 2000 4000 6000 8000 100000
01
02
03
04
05
06
07
08
09
1
The radius of cell (m)
Prob
abili
ty o
f han
doff
5kmh30kmh60kmh
75kmh90kmh120 kmh
(b)
Figure 7 The handoff probability and the average time with various cell radius and user speeds
rules And then the inference engine will make a fuzzydecision according to the mapping values But as mentionedabove the output of the fuzzy inference module is still afuzzy set and has no effects on the controlled object directlyTherefore a ldquotranslationrdquo process is needed to transfer a fuzzyvalue to a precise one119891
0This step is called ldquoreversal fuzzifierrdquo
Through comparison with this precise value we can get theaccess point candidacy value (APCV) output The goal ofthis ldquoreversal fuzzifierrdquo is to find the maximum membershipdegree function The ldquoreversal fuzzifierrdquo selects the elementwhich will make the value of membership function thehighest as the output signal And the rules could be describedas
120583 (1198910) ge 120583 (119891) 119891 isin 119865 (9)
In case of nonunique maximum value of the membershipfunction the average value will be taken as follows
1198910=
1
119873
119873
sum
119894=1
119891119894 120583 (119891
119894) ge 120583 (119891) (10)
In different networks different fuzzy inference modelsare adopted The membership function is designed to ensurecomparability among different networks at the same timeAt last in order to complete the behavior of handoff thesystem will select the highest APCV network to accessThereare three parameters that need to be detected in the fuzzymodule RSSI data rate and interference ratio Gaussian and119878-scheme distribution are chosen as the fuzzy membershipbecause of their great performance with respect to real-timecontrol
Figure 9 shows the WLAN membership functions Forexample RSSI stands for the received signal strength indi-cator Therefore we make such an assumption to describe its
knowledge rules (as shown in Figure 9(c)) (i) the member-ship function is 119878-scheme distributionwhenRSSI is in a giveninterval from minus95 dBm to minus80 dBm (ii) if RSSI belongs to theinterval [minus90 dBmminus70 dBm] then themembership functionfollows Gaussian distribution and (iii) once the value ofRSSI is higher than minus80 dBm the membership function is119878-scheme distribution as well The fuzzy set of RSSI the setof data rate and the fuzzy set of interference ratio in WLANare [good normal weak] with value set [3 2 1] Hence theoutput fuzzy sets are mapped as
119878 =
3
sum
119894=1
119872119894119896 (11)
Take a regular WLAN network as an example if RSSI isweak and data rate is weak and interference ratio is weakthen 119878 is 3 For the WLAN network each input has threefuzzy sequences so there are 27 fuzzy rules in the knowledgerules Due to the sensitivity to UE speed in V2V network5 fuzzy variable sequences are used (45 fuzzy rules) in theknowledge rules The input parameters in the fuzzy logicmodule are mapped into different fuzzy sets and the outputwould be acquired utilizing the maximum membershipdegree function method The fuzzy logic outputs decisionvalue indicates that the stable degree of difference betweentwo network decisions is lower than the threshold level thusthe connection with the source network cell will be held onOtherwise the handoff executionwill be done if the hysteresistime is longer than a certain value that is
VALUE TARGET minus VALUE SOURCE ge Hysteresis
VALUE SOURCE le Threshold(12)
6 Chinese Journal of Engineering
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The s
treng
th o
f rec
eive
d sig
nal
from
targ
et ce
ll (d
Bm)
The s
treng
th o
f rec
eive
d sig
nal
from
sour
ce ce
ll (d
Bm)
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The location of vehicle (km)
The location of vehicle (km)
(a) The strength of received signal from source and target cell
06 07 08 09 1 11 12 13 140
01
02
03
04
05
06
07
08
09
1
The location of vehicle (km)
The p
roba
bilit
y of
han
doff
The theoretical handoff point
The probability of connectingwith the target cell
The probability of connectingwith the source cell
(b) The probability of handoff
06 07 08 09 1 11 12 13 1401
02
03
04
05
06
07
08
09
10
The location of vehicle (km)
Prob
abili
ty o
f han
doff The actual handoff location point
(c) The actual handoff location point
20 40 60 80 100 120 140 160 1800
005
01
015
02
025
03
035
04
The speed of mobile terminal (kmh)
Ping
-pon
g ha
ndoff
ratio
Proposed algorithmExisting RSSI algorithm(d) The performance comparison
Figure 8 The horizontal handoff simulation results
33 Channel Model This subsection will adopt the COST 231Hata channel model to realize the real-time simulation [1718] which is depicted as
PL (dB) = (397 minus 701 log10 (ℎBS)) log10 (119889
1000
)
+ 429 + (3654 minus 12ℎMS) log10 (119891119888)
minus 1298 log10(ℎBS) + 074ℎMS + 119862
(13)
where ℎBS and ℎMS represent the height of BS and MSrespectively 119889 means the distance between BS and MS 119891119888denotes the center frequency and 119862 is a constant
The channel impulse response (CIR) from transmitterantenna element 119904 to receiver element 119906 for cluster 119899 isexpressed as
is the cross polarization power ratio in linear scale V is themoving speed of user and 120579
119899119898AoA and 120579119899119898AoD mean the
angle of arrival (AoA) and angle of departure (AoD) with the119898th subpath in 119899th path respectively Assume that there are6 paths and each path includes 20 subpaths
4 The Simulation and Validation
To evaluate the realistic performance of our proposed algo-rithm the simulation model scenarios and various parame-ters are developed using MATLAB The performance of hor-izontal and vertical handoff algorithms is tested separately
Figure 7 compares the handoff probability and averagedwell time with different moving speed and cellular radius Itis found that the expansion of cellular radius and decreasingvehicle moving speed will lead to increase of average dwelltime and channel holding time In particular the relationshipbetween the coverage radius and dwell time is obviouslylinear in the low vehicle speed regions On the contrary withthe increase of vehicle speed the handoff probability willcontinue to decrease
The horizontal handoff is simulated using the followingparameters (as shown in Figure 5) the point coordinates of
source BS and target BS are [0 50] and [2000 50] whichmeans that the distance between the source BS and targetBS is 2000m and the distance between road and BS is50m the vehicle speed is 80 kmh and is moving from [0 0]to [2000 0] the heights of BS and MS are 30m and 1mseparately the transmitted power is 44 dBm and the standarddeviation of channel shadow fading is 8 dB
Figure 8 shows the simulation results of horizontal hand-off It is clearly seen that when the vehiclemoves from [500 0]to [1500 0] the strength of received signal from source BSis reducing while the signal level from target BS is growingMeanwhile the handoff probability shows a similar trendand the handoff should occur at the points near [1000 0]according to the theoretical judgment of holding probabilityIt can also be observed that the handoff happens at the peakof the curve namely the coordinate of the vehicle is [1055 0]There is a little far distance from the actual handoff locationto the theoretical one due to the hysteresis effect of dwelltime on the joint handoff algorithm As we all know theexisting RSSI algorithm depends on the decisions of priorityon the mobile node In particular the priority is dividedinto low priority and high priority The low priority whichis below certain level estimate receives signal strength from
8 Chinese Journal of Engineering
0 50 100 150 200 250 300 350 400minus105
minus100
minus95
minus90
minus85
minus80
minus75
minus70
The simulation time (s)
RSS
(dBm
)The received signal strength
LTEWLAN
(a) RSS
0 50 100 150 200 250 3000
01
02
03
04
05
06
07
08
09
1
The simulation time (s)
Mar
k of
net
wor
k
LTE
WLANThe state of handoff
(b) The state of handoff
50 100 150 200 250 300 3500
1
2
3
4
5
6
7
8
9
10
The simulation time (s)
APC
V
LTEWLANV2V
V2VWLAN
LTE
(c) APCV and state of handoff
1 15 2 25 3 35 402
03
04
05
06
07
08
09
1
Request (MBs)
Thro
ughp
ut
V2VWLANLTE
(d) Throughput versus request
Figure 10 The vertical handoff simulation results
another cell before it starts handoff in the cell ThereforeFigure 8(d) presents a performance comparison between theexisting RSSI horizontal handoff algorithm and our proposedschemes It is clear that no matter how much the speedof users is the ping-pong handoff ratio of existing RSSIhorizontal handoff algorithm is higher than the proposedschemes It proves that ping-pong effect can be avoidedeffectively by using dwell timer and hysteresis parameters
The vertical handoff simulation is divided into two partsone focuses on testing whether the handoff could be appliedautomatically and another part analyzes the performanceof the proposed algorithm Firstly assume a car is movingfrom the initial points [0 minus50] to [2500 minus50] at the speed of45 kmh and the coordinates of WLAN-BS and LTE-eNodeare [500 0] and [2200 0] respectively (as shown in Figure 6)
Figure 10(a) shows the alternative trend of the RSS during thewhole movement and Figure 10(b) shows the actual handoffresults It is not difficult to find such a handoff process thecar accessed LTE at first and then changed the network toWLAN because it ran close to theWLAN-BS when it left thecoverage of WLAN the handoff happened again to replaceWLAN by LTE signals Obviously because of different pathloss and shadow fading due to different center frequencythe vehicle utilizes LTE network to communicate and onlyconnects to WLAN network when the mobile terminal isvery close to the BS of WLAN Secondly at the performancesimulation parts similar to the functional simulation thecoordinates of the car and the BS ofWLAN and LTE networkare the same as the former onesThe biggest difference settingbetween the function and the performance simulation part
Chinese Journal of Engineering 9
0 50 100 150 200025
03
035
04
045
05
055
06
065
07
Simulation time (s)
Load
bal
anci
ng in
dex
TTSProposed methods
(a)
50 100 150 200 250 300 350 400 450 5000
005
01
015
02
025
03
035
04
045
The number of users
Han
doff
call
drop
ping
pro
babi
lity
Proposed methodsTTS
(b)
Figure 11 The performance comparison in different methods
is that it is assumed that another car is moving from points[250 minus50] to points [minus250 minus50] at the same speed with thetarget car
As it is observed in Figure 10(c) at the beginning thecar utilized V2V network to communicate and after then thehandoff process changes the access network from WLAN toLTE according to the rules for choosing the highest valueof APCV (as shown by black lines) among them as accessnetwork The handoff process is consistent with the previoussimulation except that it accessed V2V network initially Bycomparing the throughput of different networks (as shown inFigure 10(d)) it could be found that the network performanceof LTE is the best then there isWLAN and V2V is the worstFrom the aspects of vertical handoff the triangle and trape-zoid shapes (TTS) are always chosen as fuzzy membershipfunctions in many engineering applications Moreover theload balancing degree and handoff call dropping probabilityare used to evaluate the performance of handoff Therefore
after comparison from Figure 11(a) we can find that the loadbalancing degree of our proposed schemes is higher than theTTS In addition as illustrated in Figure 11(b) call droppingprobability is lower in our proposed schemes than that ofTTS especially in the larger number of users conditionsIn general the simulation results validate that our proposedscheme outperforms the existing vertical handoff algorithm
5 Conclusion
This paper proposed a dynamic handoff algorithm in vehiclesnetworks to solve the conflicts between BS and MS such ascall blocking call dropping and ping-pong phenomenonSimulation results validate that the presented joint handoffalgorithm can effectively address the handoff problem amongvarious access networks by taking into account differentscenarios different vehicle speeds and channel conditions
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This research is supported in part by China Impor-tant National Science and Technology Specific Projects(no 2013ZX03001020-002) and by National Key Technol-ogy Research and Development Program of China (no2012BAF14B01) and by National Natural Science Foundationof China (no 61171105 and no 61322110) and by 863 ProgramProject (no 2015AA01A703) and Doctor Funding Program(no 201300051100013)
References
[1] F Siddiqui and S Zeadally ldquoMobility management acrosshybrid wireless networks trends and challengesrdquo IEEE Com-puter Communications vol 11 no 5 pp 45ndash54 2005
[2] N Nasser A Hasswa and H Hassanein ldquoHandoffs in fourthgeneration heterogeneous networksrdquo IEEE CommunicationsMagazine vol 44 no 10 pp 96ndash103 2006
[3] M Ylianttila J Makeela and PMahonen ldquoVehicular telematicsover heterogeneous wireless networks a surveyrdquo ComputerCommunications vol 33 no 7 pp 775ndash793 2010
[4] X Liu L-G Jiang andCHe ldquoVertical handoff algorithmbasedon fuzzy logic in aid of pre-decision methodrdquo Acta ElectronicaSinica vol 35 no 10 pp 1989ndash1993 2007
[5] Y B Kang K Xu Y L Shen et al ldquoOptimal distributed verticalhandoff strategies in vehicular heterogeneous networksrdquo IEEEJournal on Communications supplement 1 pp 69ndash73 2009
[6] D Ma and M Ma ldquoA QoS-based vertical handoff scheme forinterworking of WLAN and WiMAXrdquo in Proceedings of theIEEEGlobal Telecommunications Conference (GLOBECOM rsquo09)pp 1ndash6 Honolulu Hawaii USA December 2009
[7] M Liu Z-C Li X-B Guo and K Zheng ldquoA movementtrend based self-adaptive vertical handoff algorithm and itsperformance evaluationrdquo Chinese Journal of Computers vol 31no 1 pp 112ndash119 2008
[8] Y Li M Li B Cao Y Wang and W Liu ldquoDynamic optimiza-tion of handover parameters adjustment for conflict avoidance
10 Chinese Journal of Engineering
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
Figure 7 The handoff probability and the average time with various cell radius and user speeds
rules And then the inference engine will make a fuzzydecision according to the mapping values But as mentionedabove the output of the fuzzy inference module is still afuzzy set and has no effects on the controlled object directlyTherefore a ldquotranslationrdquo process is needed to transfer a fuzzyvalue to a precise one119891
0This step is called ldquoreversal fuzzifierrdquo
Through comparison with this precise value we can get theaccess point candidacy value (APCV) output The goal ofthis ldquoreversal fuzzifierrdquo is to find the maximum membershipdegree function The ldquoreversal fuzzifierrdquo selects the elementwhich will make the value of membership function thehighest as the output signal And the rules could be describedas
120583 (1198910) ge 120583 (119891) 119891 isin 119865 (9)
In case of nonunique maximum value of the membershipfunction the average value will be taken as follows
1198910=
1
119873
119873
sum
119894=1
119891119894 120583 (119891
119894) ge 120583 (119891) (10)
In different networks different fuzzy inference modelsare adopted The membership function is designed to ensurecomparability among different networks at the same timeAt last in order to complete the behavior of handoff thesystem will select the highest APCV network to accessThereare three parameters that need to be detected in the fuzzymodule RSSI data rate and interference ratio Gaussian and119878-scheme distribution are chosen as the fuzzy membershipbecause of their great performance with respect to real-timecontrol
Figure 9 shows the WLAN membership functions Forexample RSSI stands for the received signal strength indi-cator Therefore we make such an assumption to describe its
knowledge rules (as shown in Figure 9(c)) (i) the member-ship function is 119878-scheme distributionwhenRSSI is in a giveninterval from minus95 dBm to minus80 dBm (ii) if RSSI belongs to theinterval [minus90 dBmminus70 dBm] then themembership functionfollows Gaussian distribution and (iii) once the value ofRSSI is higher than minus80 dBm the membership function is119878-scheme distribution as well The fuzzy set of RSSI the setof data rate and the fuzzy set of interference ratio in WLANare [good normal weak] with value set [3 2 1] Hence theoutput fuzzy sets are mapped as
119878 =
3
sum
119894=1
119872119894119896 (11)
Take a regular WLAN network as an example if RSSI isweak and data rate is weak and interference ratio is weakthen 119878 is 3 For the WLAN network each input has threefuzzy sequences so there are 27 fuzzy rules in the knowledgerules Due to the sensitivity to UE speed in V2V network5 fuzzy variable sequences are used (45 fuzzy rules) in theknowledge rules The input parameters in the fuzzy logicmodule are mapped into different fuzzy sets and the outputwould be acquired utilizing the maximum membershipdegree function method The fuzzy logic outputs decisionvalue indicates that the stable degree of difference betweentwo network decisions is lower than the threshold level thusthe connection with the source network cell will be held onOtherwise the handoff executionwill be done if the hysteresistime is longer than a certain value that is
VALUE TARGET minus VALUE SOURCE ge Hysteresis
VALUE SOURCE le Threshold(12)
6 Chinese Journal of Engineering
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The s
treng
th o
f rec
eive
d sig
nal
from
targ
et ce
ll (d
Bm)
The s
treng
th o
f rec
eive
d sig
nal
from
sour
ce ce
ll (d
Bm)
05 06 07 08 09 1 11 12 13 14 15minus110
minus100
minus90
minus80
minus70
The location of vehicle (km)
The location of vehicle (km)
(a) The strength of received signal from source and target cell
06 07 08 09 1 11 12 13 140
01
02
03
04
05
06
07
08
09
1
The location of vehicle (km)
The p
roba
bilit
y of
han
doff
The theoretical handoff point
The probability of connectingwith the target cell
The probability of connectingwith the source cell
(b) The probability of handoff
06 07 08 09 1 11 12 13 1401
02
03
04
05
06
07
08
09
10
The location of vehicle (km)
Prob
abili
ty o
f han
doff The actual handoff location point
(c) The actual handoff location point
20 40 60 80 100 120 140 160 1800
005
01
015
02
025
03
035
04
The speed of mobile terminal (kmh)
Ping
-pon
g ha
ndoff
ratio
Proposed algorithmExisting RSSI algorithm(d) The performance comparison
Figure 8 The horizontal handoff simulation results
33 Channel Model This subsection will adopt the COST 231Hata channel model to realize the real-time simulation [1718] which is depicted as
PL (dB) = (397 minus 701 log10 (ℎBS)) log10 (119889
1000
)
+ 429 + (3654 minus 12ℎMS) log10 (119891119888)
minus 1298 log10(ℎBS) + 074ℎMS + 119862
(13)
where ℎBS and ℎMS represent the height of BS and MSrespectively 119889 means the distance between BS and MS 119891119888denotes the center frequency and 119862 is a constant
The channel impulse response (CIR) from transmitterantenna element 119904 to receiver element 119906 for cluster 119899 isexpressed as
is the cross polarization power ratio in linear scale V is themoving speed of user and 120579
119899119898AoA and 120579119899119898AoD mean the
angle of arrival (AoA) and angle of departure (AoD) with the119898th subpath in 119899th path respectively Assume that there are6 paths and each path includes 20 subpaths
4 The Simulation and Validation
To evaluate the realistic performance of our proposed algo-rithm the simulation model scenarios and various parame-ters are developed using MATLAB The performance of hor-izontal and vertical handoff algorithms is tested separately
Figure 7 compares the handoff probability and averagedwell time with different moving speed and cellular radius Itis found that the expansion of cellular radius and decreasingvehicle moving speed will lead to increase of average dwelltime and channel holding time In particular the relationshipbetween the coverage radius and dwell time is obviouslylinear in the low vehicle speed regions On the contrary withthe increase of vehicle speed the handoff probability willcontinue to decrease
The horizontal handoff is simulated using the followingparameters (as shown in Figure 5) the point coordinates of
source BS and target BS are [0 50] and [2000 50] whichmeans that the distance between the source BS and targetBS is 2000m and the distance between road and BS is50m the vehicle speed is 80 kmh and is moving from [0 0]to [2000 0] the heights of BS and MS are 30m and 1mseparately the transmitted power is 44 dBm and the standarddeviation of channel shadow fading is 8 dB
Figure 8 shows the simulation results of horizontal hand-off It is clearly seen that when the vehiclemoves from [500 0]to [1500 0] the strength of received signal from source BSis reducing while the signal level from target BS is growingMeanwhile the handoff probability shows a similar trendand the handoff should occur at the points near [1000 0]according to the theoretical judgment of holding probabilityIt can also be observed that the handoff happens at the peakof the curve namely the coordinate of the vehicle is [1055 0]There is a little far distance from the actual handoff locationto the theoretical one due to the hysteresis effect of dwelltime on the joint handoff algorithm As we all know theexisting RSSI algorithm depends on the decisions of priorityon the mobile node In particular the priority is dividedinto low priority and high priority The low priority whichis below certain level estimate receives signal strength from
8 Chinese Journal of Engineering
0 50 100 150 200 250 300 350 400minus105
minus100
minus95
minus90
minus85
minus80
minus75
minus70
The simulation time (s)
RSS
(dBm
)The received signal strength
LTEWLAN
(a) RSS
0 50 100 150 200 250 3000
01
02
03
04
05
06
07
08
09
1
The simulation time (s)
Mar
k of
net
wor
k
LTE
WLANThe state of handoff
(b) The state of handoff
50 100 150 200 250 300 3500
1
2
3
4
5
6
7
8
9
10
The simulation time (s)
APC
V
LTEWLANV2V
V2VWLAN
LTE
(c) APCV and state of handoff
1 15 2 25 3 35 402
03
04
05
06
07
08
09
1
Request (MBs)
Thro
ughp
ut
V2VWLANLTE
(d) Throughput versus request
Figure 10 The vertical handoff simulation results
another cell before it starts handoff in the cell ThereforeFigure 8(d) presents a performance comparison between theexisting RSSI horizontal handoff algorithm and our proposedschemes It is clear that no matter how much the speedof users is the ping-pong handoff ratio of existing RSSIhorizontal handoff algorithm is higher than the proposedschemes It proves that ping-pong effect can be avoidedeffectively by using dwell timer and hysteresis parameters
The vertical handoff simulation is divided into two partsone focuses on testing whether the handoff could be appliedautomatically and another part analyzes the performanceof the proposed algorithm Firstly assume a car is movingfrom the initial points [0 minus50] to [2500 minus50] at the speed of45 kmh and the coordinates of WLAN-BS and LTE-eNodeare [500 0] and [2200 0] respectively (as shown in Figure 6)
Figure 10(a) shows the alternative trend of the RSS during thewhole movement and Figure 10(b) shows the actual handoffresults It is not difficult to find such a handoff process thecar accessed LTE at first and then changed the network toWLAN because it ran close to theWLAN-BS when it left thecoverage of WLAN the handoff happened again to replaceWLAN by LTE signals Obviously because of different pathloss and shadow fading due to different center frequencythe vehicle utilizes LTE network to communicate and onlyconnects to WLAN network when the mobile terminal isvery close to the BS of WLAN Secondly at the performancesimulation parts similar to the functional simulation thecoordinates of the car and the BS ofWLAN and LTE networkare the same as the former onesThe biggest difference settingbetween the function and the performance simulation part
Chinese Journal of Engineering 9
0 50 100 150 200025
03
035
04
045
05
055
06
065
07
Simulation time (s)
Load
bal
anci
ng in
dex
TTSProposed methods
(a)
50 100 150 200 250 300 350 400 450 5000
005
01
015
02
025
03
035
04
045
The number of users
Han
doff
call
drop
ping
pro
babi
lity
Proposed methodsTTS
(b)
Figure 11 The performance comparison in different methods
is that it is assumed that another car is moving from points[250 minus50] to points [minus250 minus50] at the same speed with thetarget car
As it is observed in Figure 10(c) at the beginning thecar utilized V2V network to communicate and after then thehandoff process changes the access network from WLAN toLTE according to the rules for choosing the highest valueof APCV (as shown by black lines) among them as accessnetwork The handoff process is consistent with the previoussimulation except that it accessed V2V network initially Bycomparing the throughput of different networks (as shown inFigure 10(d)) it could be found that the network performanceof LTE is the best then there isWLAN and V2V is the worstFrom the aspects of vertical handoff the triangle and trape-zoid shapes (TTS) are always chosen as fuzzy membershipfunctions in many engineering applications Moreover theload balancing degree and handoff call dropping probabilityare used to evaluate the performance of handoff Therefore
after comparison from Figure 11(a) we can find that the loadbalancing degree of our proposed schemes is higher than theTTS In addition as illustrated in Figure 11(b) call droppingprobability is lower in our proposed schemes than that ofTTS especially in the larger number of users conditionsIn general the simulation results validate that our proposedscheme outperforms the existing vertical handoff algorithm
5 Conclusion
This paper proposed a dynamic handoff algorithm in vehiclesnetworks to solve the conflicts between BS and MS such ascall blocking call dropping and ping-pong phenomenonSimulation results validate that the presented joint handoffalgorithm can effectively address the handoff problem amongvarious access networks by taking into account differentscenarios different vehicle speeds and channel conditions
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This research is supported in part by China Impor-tant National Science and Technology Specific Projects(no 2013ZX03001020-002) and by National Key Technol-ogy Research and Development Program of China (no2012BAF14B01) and by National Natural Science Foundationof China (no 61171105 and no 61322110) and by 863 ProgramProject (no 2015AA01A703) and Doctor Funding Program(no 201300051100013)
References
[1] F Siddiqui and S Zeadally ldquoMobility management acrosshybrid wireless networks trends and challengesrdquo IEEE Com-puter Communications vol 11 no 5 pp 45ndash54 2005
[2] N Nasser A Hasswa and H Hassanein ldquoHandoffs in fourthgeneration heterogeneous networksrdquo IEEE CommunicationsMagazine vol 44 no 10 pp 96ndash103 2006
[3] M Ylianttila J Makeela and PMahonen ldquoVehicular telematicsover heterogeneous wireless networks a surveyrdquo ComputerCommunications vol 33 no 7 pp 775ndash793 2010
[4] X Liu L-G Jiang andCHe ldquoVertical handoff algorithmbasedon fuzzy logic in aid of pre-decision methodrdquo Acta ElectronicaSinica vol 35 no 10 pp 1989ndash1993 2007
[5] Y B Kang K Xu Y L Shen et al ldquoOptimal distributed verticalhandoff strategies in vehicular heterogeneous networksrdquo IEEEJournal on Communications supplement 1 pp 69ndash73 2009
[6] D Ma and M Ma ldquoA QoS-based vertical handoff scheme forinterworking of WLAN and WiMAXrdquo in Proceedings of theIEEEGlobal Telecommunications Conference (GLOBECOM rsquo09)pp 1ndash6 Honolulu Hawaii USA December 2009
[7] M Liu Z-C Li X-B Guo and K Zheng ldquoA movementtrend based self-adaptive vertical handoff algorithm and itsperformance evaluationrdquo Chinese Journal of Computers vol 31no 1 pp 112ndash119 2008
[8] Y Li M Li B Cao Y Wang and W Liu ldquoDynamic optimiza-tion of handover parameters adjustment for conflict avoidance
10 Chinese Journal of Engineering
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
(a) The strength of received signal from source and target cell
06 07 08 09 1 11 12 13 140
01
02
03
04
05
06
07
08
09
1
The location of vehicle (km)
The p
roba
bilit
y of
han
doff
The theoretical handoff point
The probability of connectingwith the target cell
The probability of connectingwith the source cell
(b) The probability of handoff
06 07 08 09 1 11 12 13 1401
02
03
04
05
06
07
08
09
10
The location of vehicle (km)
Prob
abili
ty o
f han
doff The actual handoff location point
(c) The actual handoff location point
20 40 60 80 100 120 140 160 1800
005
01
015
02
025
03
035
04
The speed of mobile terminal (kmh)
Ping
-pon
g ha
ndoff
ratio
Proposed algorithmExisting RSSI algorithm(d) The performance comparison
Figure 8 The horizontal handoff simulation results
33 Channel Model This subsection will adopt the COST 231Hata channel model to realize the real-time simulation [1718] which is depicted as
PL (dB) = (397 minus 701 log10 (ℎBS)) log10 (119889
1000
)
+ 429 + (3654 minus 12ℎMS) log10 (119891119888)
minus 1298 log10(ℎBS) + 074ℎMS + 119862
(13)
where ℎBS and ℎMS represent the height of BS and MSrespectively 119889 means the distance between BS and MS 119891119888denotes the center frequency and 119862 is a constant
The channel impulse response (CIR) from transmitterantenna element 119904 to receiver element 119906 for cluster 119899 isexpressed as
is the cross polarization power ratio in linear scale V is themoving speed of user and 120579
119899119898AoA and 120579119899119898AoD mean the
angle of arrival (AoA) and angle of departure (AoD) with the119898th subpath in 119899th path respectively Assume that there are6 paths and each path includes 20 subpaths
4 The Simulation and Validation
To evaluate the realistic performance of our proposed algo-rithm the simulation model scenarios and various parame-ters are developed using MATLAB The performance of hor-izontal and vertical handoff algorithms is tested separately
Figure 7 compares the handoff probability and averagedwell time with different moving speed and cellular radius Itis found that the expansion of cellular radius and decreasingvehicle moving speed will lead to increase of average dwelltime and channel holding time In particular the relationshipbetween the coverage radius and dwell time is obviouslylinear in the low vehicle speed regions On the contrary withthe increase of vehicle speed the handoff probability willcontinue to decrease
The horizontal handoff is simulated using the followingparameters (as shown in Figure 5) the point coordinates of
source BS and target BS are [0 50] and [2000 50] whichmeans that the distance between the source BS and targetBS is 2000m and the distance between road and BS is50m the vehicle speed is 80 kmh and is moving from [0 0]to [2000 0] the heights of BS and MS are 30m and 1mseparately the transmitted power is 44 dBm and the standarddeviation of channel shadow fading is 8 dB
Figure 8 shows the simulation results of horizontal hand-off It is clearly seen that when the vehiclemoves from [500 0]to [1500 0] the strength of received signal from source BSis reducing while the signal level from target BS is growingMeanwhile the handoff probability shows a similar trendand the handoff should occur at the points near [1000 0]according to the theoretical judgment of holding probabilityIt can also be observed that the handoff happens at the peakof the curve namely the coordinate of the vehicle is [1055 0]There is a little far distance from the actual handoff locationto the theoretical one due to the hysteresis effect of dwelltime on the joint handoff algorithm As we all know theexisting RSSI algorithm depends on the decisions of priorityon the mobile node In particular the priority is dividedinto low priority and high priority The low priority whichis below certain level estimate receives signal strength from
8 Chinese Journal of Engineering
0 50 100 150 200 250 300 350 400minus105
minus100
minus95
minus90
minus85
minus80
minus75
minus70
The simulation time (s)
RSS
(dBm
)The received signal strength
LTEWLAN
(a) RSS
0 50 100 150 200 250 3000
01
02
03
04
05
06
07
08
09
1
The simulation time (s)
Mar
k of
net
wor
k
LTE
WLANThe state of handoff
(b) The state of handoff
50 100 150 200 250 300 3500
1
2
3
4
5
6
7
8
9
10
The simulation time (s)
APC
V
LTEWLANV2V
V2VWLAN
LTE
(c) APCV and state of handoff
1 15 2 25 3 35 402
03
04
05
06
07
08
09
1
Request (MBs)
Thro
ughp
ut
V2VWLANLTE
(d) Throughput versus request
Figure 10 The vertical handoff simulation results
another cell before it starts handoff in the cell ThereforeFigure 8(d) presents a performance comparison between theexisting RSSI horizontal handoff algorithm and our proposedschemes It is clear that no matter how much the speedof users is the ping-pong handoff ratio of existing RSSIhorizontal handoff algorithm is higher than the proposedschemes It proves that ping-pong effect can be avoidedeffectively by using dwell timer and hysteresis parameters
The vertical handoff simulation is divided into two partsone focuses on testing whether the handoff could be appliedautomatically and another part analyzes the performanceof the proposed algorithm Firstly assume a car is movingfrom the initial points [0 minus50] to [2500 minus50] at the speed of45 kmh and the coordinates of WLAN-BS and LTE-eNodeare [500 0] and [2200 0] respectively (as shown in Figure 6)
Figure 10(a) shows the alternative trend of the RSS during thewhole movement and Figure 10(b) shows the actual handoffresults It is not difficult to find such a handoff process thecar accessed LTE at first and then changed the network toWLAN because it ran close to theWLAN-BS when it left thecoverage of WLAN the handoff happened again to replaceWLAN by LTE signals Obviously because of different pathloss and shadow fading due to different center frequencythe vehicle utilizes LTE network to communicate and onlyconnects to WLAN network when the mobile terminal isvery close to the BS of WLAN Secondly at the performancesimulation parts similar to the functional simulation thecoordinates of the car and the BS ofWLAN and LTE networkare the same as the former onesThe biggest difference settingbetween the function and the performance simulation part
Chinese Journal of Engineering 9
0 50 100 150 200025
03
035
04
045
05
055
06
065
07
Simulation time (s)
Load
bal
anci
ng in
dex
TTSProposed methods
(a)
50 100 150 200 250 300 350 400 450 5000
005
01
015
02
025
03
035
04
045
The number of users
Han
doff
call
drop
ping
pro
babi
lity
Proposed methodsTTS
(b)
Figure 11 The performance comparison in different methods
is that it is assumed that another car is moving from points[250 minus50] to points [minus250 minus50] at the same speed with thetarget car
As it is observed in Figure 10(c) at the beginning thecar utilized V2V network to communicate and after then thehandoff process changes the access network from WLAN toLTE according to the rules for choosing the highest valueof APCV (as shown by black lines) among them as accessnetwork The handoff process is consistent with the previoussimulation except that it accessed V2V network initially Bycomparing the throughput of different networks (as shown inFigure 10(d)) it could be found that the network performanceof LTE is the best then there isWLAN and V2V is the worstFrom the aspects of vertical handoff the triangle and trape-zoid shapes (TTS) are always chosen as fuzzy membershipfunctions in many engineering applications Moreover theload balancing degree and handoff call dropping probabilityare used to evaluate the performance of handoff Therefore
after comparison from Figure 11(a) we can find that the loadbalancing degree of our proposed schemes is higher than theTTS In addition as illustrated in Figure 11(b) call droppingprobability is lower in our proposed schemes than that ofTTS especially in the larger number of users conditionsIn general the simulation results validate that our proposedscheme outperforms the existing vertical handoff algorithm
5 Conclusion
This paper proposed a dynamic handoff algorithm in vehiclesnetworks to solve the conflicts between BS and MS such ascall blocking call dropping and ping-pong phenomenonSimulation results validate that the presented joint handoffalgorithm can effectively address the handoff problem amongvarious access networks by taking into account differentscenarios different vehicle speeds and channel conditions
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This research is supported in part by China Impor-tant National Science and Technology Specific Projects(no 2013ZX03001020-002) and by National Key Technol-ogy Research and Development Program of China (no2012BAF14B01) and by National Natural Science Foundationof China (no 61171105 and no 61322110) and by 863 ProgramProject (no 2015AA01A703) and Doctor Funding Program(no 201300051100013)
References
[1] F Siddiqui and S Zeadally ldquoMobility management acrosshybrid wireless networks trends and challengesrdquo IEEE Com-puter Communications vol 11 no 5 pp 45ndash54 2005
[2] N Nasser A Hasswa and H Hassanein ldquoHandoffs in fourthgeneration heterogeneous networksrdquo IEEE CommunicationsMagazine vol 44 no 10 pp 96ndash103 2006
[3] M Ylianttila J Makeela and PMahonen ldquoVehicular telematicsover heterogeneous wireless networks a surveyrdquo ComputerCommunications vol 33 no 7 pp 775ndash793 2010
[4] X Liu L-G Jiang andCHe ldquoVertical handoff algorithmbasedon fuzzy logic in aid of pre-decision methodrdquo Acta ElectronicaSinica vol 35 no 10 pp 1989ndash1993 2007
[5] Y B Kang K Xu Y L Shen et al ldquoOptimal distributed verticalhandoff strategies in vehicular heterogeneous networksrdquo IEEEJournal on Communications supplement 1 pp 69ndash73 2009
[6] D Ma and M Ma ldquoA QoS-based vertical handoff scheme forinterworking of WLAN and WiMAXrdquo in Proceedings of theIEEEGlobal Telecommunications Conference (GLOBECOM rsquo09)pp 1ndash6 Honolulu Hawaii USA December 2009
[7] M Liu Z-C Li X-B Guo and K Zheng ldquoA movementtrend based self-adaptive vertical handoff algorithm and itsperformance evaluationrdquo Chinese Journal of Computers vol 31no 1 pp 112ndash119 2008
[8] Y Li M Li B Cao Y Wang and W Liu ldquoDynamic optimiza-tion of handover parameters adjustment for conflict avoidance
10 Chinese Journal of Engineering
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
is the cross polarization power ratio in linear scale V is themoving speed of user and 120579
119899119898AoA and 120579119899119898AoD mean the
angle of arrival (AoA) and angle of departure (AoD) with the119898th subpath in 119899th path respectively Assume that there are6 paths and each path includes 20 subpaths
4 The Simulation and Validation
To evaluate the realistic performance of our proposed algo-rithm the simulation model scenarios and various parame-ters are developed using MATLAB The performance of hor-izontal and vertical handoff algorithms is tested separately
Figure 7 compares the handoff probability and averagedwell time with different moving speed and cellular radius Itis found that the expansion of cellular radius and decreasingvehicle moving speed will lead to increase of average dwelltime and channel holding time In particular the relationshipbetween the coverage radius and dwell time is obviouslylinear in the low vehicle speed regions On the contrary withthe increase of vehicle speed the handoff probability willcontinue to decrease
The horizontal handoff is simulated using the followingparameters (as shown in Figure 5) the point coordinates of
source BS and target BS are [0 50] and [2000 50] whichmeans that the distance between the source BS and targetBS is 2000m and the distance between road and BS is50m the vehicle speed is 80 kmh and is moving from [0 0]to [2000 0] the heights of BS and MS are 30m and 1mseparately the transmitted power is 44 dBm and the standarddeviation of channel shadow fading is 8 dB
Figure 8 shows the simulation results of horizontal hand-off It is clearly seen that when the vehiclemoves from [500 0]to [1500 0] the strength of received signal from source BSis reducing while the signal level from target BS is growingMeanwhile the handoff probability shows a similar trendand the handoff should occur at the points near [1000 0]according to the theoretical judgment of holding probabilityIt can also be observed that the handoff happens at the peakof the curve namely the coordinate of the vehicle is [1055 0]There is a little far distance from the actual handoff locationto the theoretical one due to the hysteresis effect of dwelltime on the joint handoff algorithm As we all know theexisting RSSI algorithm depends on the decisions of priorityon the mobile node In particular the priority is dividedinto low priority and high priority The low priority whichis below certain level estimate receives signal strength from
8 Chinese Journal of Engineering
0 50 100 150 200 250 300 350 400minus105
minus100
minus95
minus90
minus85
minus80
minus75
minus70
The simulation time (s)
RSS
(dBm
)The received signal strength
LTEWLAN
(a) RSS
0 50 100 150 200 250 3000
01
02
03
04
05
06
07
08
09
1
The simulation time (s)
Mar
k of
net
wor
k
LTE
WLANThe state of handoff
(b) The state of handoff
50 100 150 200 250 300 3500
1
2
3
4
5
6
7
8
9
10
The simulation time (s)
APC
V
LTEWLANV2V
V2VWLAN
LTE
(c) APCV and state of handoff
1 15 2 25 3 35 402
03
04
05
06
07
08
09
1
Request (MBs)
Thro
ughp
ut
V2VWLANLTE
(d) Throughput versus request
Figure 10 The vertical handoff simulation results
another cell before it starts handoff in the cell ThereforeFigure 8(d) presents a performance comparison between theexisting RSSI horizontal handoff algorithm and our proposedschemes It is clear that no matter how much the speedof users is the ping-pong handoff ratio of existing RSSIhorizontal handoff algorithm is higher than the proposedschemes It proves that ping-pong effect can be avoidedeffectively by using dwell timer and hysteresis parameters
The vertical handoff simulation is divided into two partsone focuses on testing whether the handoff could be appliedautomatically and another part analyzes the performanceof the proposed algorithm Firstly assume a car is movingfrom the initial points [0 minus50] to [2500 minus50] at the speed of45 kmh and the coordinates of WLAN-BS and LTE-eNodeare [500 0] and [2200 0] respectively (as shown in Figure 6)
Figure 10(a) shows the alternative trend of the RSS during thewhole movement and Figure 10(b) shows the actual handoffresults It is not difficult to find such a handoff process thecar accessed LTE at first and then changed the network toWLAN because it ran close to theWLAN-BS when it left thecoverage of WLAN the handoff happened again to replaceWLAN by LTE signals Obviously because of different pathloss and shadow fading due to different center frequencythe vehicle utilizes LTE network to communicate and onlyconnects to WLAN network when the mobile terminal isvery close to the BS of WLAN Secondly at the performancesimulation parts similar to the functional simulation thecoordinates of the car and the BS ofWLAN and LTE networkare the same as the former onesThe biggest difference settingbetween the function and the performance simulation part
Chinese Journal of Engineering 9
0 50 100 150 200025
03
035
04
045
05
055
06
065
07
Simulation time (s)
Load
bal
anci
ng in
dex
TTSProposed methods
(a)
50 100 150 200 250 300 350 400 450 5000
005
01
015
02
025
03
035
04
045
The number of users
Han
doff
call
drop
ping
pro
babi
lity
Proposed methodsTTS
(b)
Figure 11 The performance comparison in different methods
is that it is assumed that another car is moving from points[250 minus50] to points [minus250 minus50] at the same speed with thetarget car
As it is observed in Figure 10(c) at the beginning thecar utilized V2V network to communicate and after then thehandoff process changes the access network from WLAN toLTE according to the rules for choosing the highest valueof APCV (as shown by black lines) among them as accessnetwork The handoff process is consistent with the previoussimulation except that it accessed V2V network initially Bycomparing the throughput of different networks (as shown inFigure 10(d)) it could be found that the network performanceof LTE is the best then there isWLAN and V2V is the worstFrom the aspects of vertical handoff the triangle and trape-zoid shapes (TTS) are always chosen as fuzzy membershipfunctions in many engineering applications Moreover theload balancing degree and handoff call dropping probabilityare used to evaluate the performance of handoff Therefore
after comparison from Figure 11(a) we can find that the loadbalancing degree of our proposed schemes is higher than theTTS In addition as illustrated in Figure 11(b) call droppingprobability is lower in our proposed schemes than that ofTTS especially in the larger number of users conditionsIn general the simulation results validate that our proposedscheme outperforms the existing vertical handoff algorithm
5 Conclusion
This paper proposed a dynamic handoff algorithm in vehiclesnetworks to solve the conflicts between BS and MS such ascall blocking call dropping and ping-pong phenomenonSimulation results validate that the presented joint handoffalgorithm can effectively address the handoff problem amongvarious access networks by taking into account differentscenarios different vehicle speeds and channel conditions
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This research is supported in part by China Impor-tant National Science and Technology Specific Projects(no 2013ZX03001020-002) and by National Key Technol-ogy Research and Development Program of China (no2012BAF14B01) and by National Natural Science Foundationof China (no 61171105 and no 61322110) and by 863 ProgramProject (no 2015AA01A703) and Doctor Funding Program(no 201300051100013)
References
[1] F Siddiqui and S Zeadally ldquoMobility management acrosshybrid wireless networks trends and challengesrdquo IEEE Com-puter Communications vol 11 no 5 pp 45ndash54 2005
[2] N Nasser A Hasswa and H Hassanein ldquoHandoffs in fourthgeneration heterogeneous networksrdquo IEEE CommunicationsMagazine vol 44 no 10 pp 96ndash103 2006
[3] M Ylianttila J Makeela and PMahonen ldquoVehicular telematicsover heterogeneous wireless networks a surveyrdquo ComputerCommunications vol 33 no 7 pp 775ndash793 2010
[4] X Liu L-G Jiang andCHe ldquoVertical handoff algorithmbasedon fuzzy logic in aid of pre-decision methodrdquo Acta ElectronicaSinica vol 35 no 10 pp 1989ndash1993 2007
[5] Y B Kang K Xu Y L Shen et al ldquoOptimal distributed verticalhandoff strategies in vehicular heterogeneous networksrdquo IEEEJournal on Communications supplement 1 pp 69ndash73 2009
[6] D Ma and M Ma ldquoA QoS-based vertical handoff scheme forinterworking of WLAN and WiMAXrdquo in Proceedings of theIEEEGlobal Telecommunications Conference (GLOBECOM rsquo09)pp 1ndash6 Honolulu Hawaii USA December 2009
[7] M Liu Z-C Li X-B Guo and K Zheng ldquoA movementtrend based self-adaptive vertical handoff algorithm and itsperformance evaluationrdquo Chinese Journal of Computers vol 31no 1 pp 112ndash119 2008
[8] Y Li M Li B Cao Y Wang and W Liu ldquoDynamic optimiza-tion of handover parameters adjustment for conflict avoidance
10 Chinese Journal of Engineering
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
another cell before it starts handoff in the cell ThereforeFigure 8(d) presents a performance comparison between theexisting RSSI horizontal handoff algorithm and our proposedschemes It is clear that no matter how much the speedof users is the ping-pong handoff ratio of existing RSSIhorizontal handoff algorithm is higher than the proposedschemes It proves that ping-pong effect can be avoidedeffectively by using dwell timer and hysteresis parameters
The vertical handoff simulation is divided into two partsone focuses on testing whether the handoff could be appliedautomatically and another part analyzes the performanceof the proposed algorithm Firstly assume a car is movingfrom the initial points [0 minus50] to [2500 minus50] at the speed of45 kmh and the coordinates of WLAN-BS and LTE-eNodeare [500 0] and [2200 0] respectively (as shown in Figure 6)
Figure 10(a) shows the alternative trend of the RSS during thewhole movement and Figure 10(b) shows the actual handoffresults It is not difficult to find such a handoff process thecar accessed LTE at first and then changed the network toWLAN because it ran close to theWLAN-BS when it left thecoverage of WLAN the handoff happened again to replaceWLAN by LTE signals Obviously because of different pathloss and shadow fading due to different center frequencythe vehicle utilizes LTE network to communicate and onlyconnects to WLAN network when the mobile terminal isvery close to the BS of WLAN Secondly at the performancesimulation parts similar to the functional simulation thecoordinates of the car and the BS ofWLAN and LTE networkare the same as the former onesThe biggest difference settingbetween the function and the performance simulation part
Chinese Journal of Engineering 9
0 50 100 150 200025
03
035
04
045
05
055
06
065
07
Simulation time (s)
Load
bal
anci
ng in
dex
TTSProposed methods
(a)
50 100 150 200 250 300 350 400 450 5000
005
01
015
02
025
03
035
04
045
The number of users
Han
doff
call
drop
ping
pro
babi
lity
Proposed methodsTTS
(b)
Figure 11 The performance comparison in different methods
is that it is assumed that another car is moving from points[250 minus50] to points [minus250 minus50] at the same speed with thetarget car
As it is observed in Figure 10(c) at the beginning thecar utilized V2V network to communicate and after then thehandoff process changes the access network from WLAN toLTE according to the rules for choosing the highest valueof APCV (as shown by black lines) among them as accessnetwork The handoff process is consistent with the previoussimulation except that it accessed V2V network initially Bycomparing the throughput of different networks (as shown inFigure 10(d)) it could be found that the network performanceof LTE is the best then there isWLAN and V2V is the worstFrom the aspects of vertical handoff the triangle and trape-zoid shapes (TTS) are always chosen as fuzzy membershipfunctions in many engineering applications Moreover theload balancing degree and handoff call dropping probabilityare used to evaluate the performance of handoff Therefore
after comparison from Figure 11(a) we can find that the loadbalancing degree of our proposed schemes is higher than theTTS In addition as illustrated in Figure 11(b) call droppingprobability is lower in our proposed schemes than that ofTTS especially in the larger number of users conditionsIn general the simulation results validate that our proposedscheme outperforms the existing vertical handoff algorithm
5 Conclusion
This paper proposed a dynamic handoff algorithm in vehiclesnetworks to solve the conflicts between BS and MS such ascall blocking call dropping and ping-pong phenomenonSimulation results validate that the presented joint handoffalgorithm can effectively address the handoff problem amongvarious access networks by taking into account differentscenarios different vehicle speeds and channel conditions
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This research is supported in part by China Impor-tant National Science and Technology Specific Projects(no 2013ZX03001020-002) and by National Key Technol-ogy Research and Development Program of China (no2012BAF14B01) and by National Natural Science Foundationof China (no 61171105 and no 61322110) and by 863 ProgramProject (no 2015AA01A703) and Doctor Funding Program(no 201300051100013)
References
[1] F Siddiqui and S Zeadally ldquoMobility management acrosshybrid wireless networks trends and challengesrdquo IEEE Com-puter Communications vol 11 no 5 pp 45ndash54 2005
[2] N Nasser A Hasswa and H Hassanein ldquoHandoffs in fourthgeneration heterogeneous networksrdquo IEEE CommunicationsMagazine vol 44 no 10 pp 96ndash103 2006
[3] M Ylianttila J Makeela and PMahonen ldquoVehicular telematicsover heterogeneous wireless networks a surveyrdquo ComputerCommunications vol 33 no 7 pp 775ndash793 2010
[4] X Liu L-G Jiang andCHe ldquoVertical handoff algorithmbasedon fuzzy logic in aid of pre-decision methodrdquo Acta ElectronicaSinica vol 35 no 10 pp 1989ndash1993 2007
[5] Y B Kang K Xu Y L Shen et al ldquoOptimal distributed verticalhandoff strategies in vehicular heterogeneous networksrdquo IEEEJournal on Communications supplement 1 pp 69ndash73 2009
[6] D Ma and M Ma ldquoA QoS-based vertical handoff scheme forinterworking of WLAN and WiMAXrdquo in Proceedings of theIEEEGlobal Telecommunications Conference (GLOBECOM rsquo09)pp 1ndash6 Honolulu Hawaii USA December 2009
[7] M Liu Z-C Li X-B Guo and K Zheng ldquoA movementtrend based self-adaptive vertical handoff algorithm and itsperformance evaluationrdquo Chinese Journal of Computers vol 31no 1 pp 112ndash119 2008
[8] Y Li M Li B Cao Y Wang and W Liu ldquoDynamic optimiza-tion of handover parameters adjustment for conflict avoidance
10 Chinese Journal of Engineering
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
Figure 11 The performance comparison in different methods
is that it is assumed that another car is moving from points[250 minus50] to points [minus250 minus50] at the same speed with thetarget car
As it is observed in Figure 10(c) at the beginning thecar utilized V2V network to communicate and after then thehandoff process changes the access network from WLAN toLTE according to the rules for choosing the highest valueof APCV (as shown by black lines) among them as accessnetwork The handoff process is consistent with the previoussimulation except that it accessed V2V network initially Bycomparing the throughput of different networks (as shown inFigure 10(d)) it could be found that the network performanceof LTE is the best then there isWLAN and V2V is the worstFrom the aspects of vertical handoff the triangle and trape-zoid shapes (TTS) are always chosen as fuzzy membershipfunctions in many engineering applications Moreover theload balancing degree and handoff call dropping probabilityare used to evaluate the performance of handoff Therefore
after comparison from Figure 11(a) we can find that the loadbalancing degree of our proposed schemes is higher than theTTS In addition as illustrated in Figure 11(b) call droppingprobability is lower in our proposed schemes than that ofTTS especially in the larger number of users conditionsIn general the simulation results validate that our proposedscheme outperforms the existing vertical handoff algorithm
5 Conclusion
This paper proposed a dynamic handoff algorithm in vehiclesnetworks to solve the conflicts between BS and MS such ascall blocking call dropping and ping-pong phenomenonSimulation results validate that the presented joint handoffalgorithm can effectively address the handoff problem amongvarious access networks by taking into account differentscenarios different vehicle speeds and channel conditions
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This research is supported in part by China Impor-tant National Science and Technology Specific Projects(no 2013ZX03001020-002) and by National Key Technol-ogy Research and Development Program of China (no2012BAF14B01) and by National Natural Science Foundationof China (no 61171105 and no 61322110) and by 863 ProgramProject (no 2015AA01A703) and Doctor Funding Program(no 201300051100013)
References
[1] F Siddiqui and S Zeadally ldquoMobility management acrosshybrid wireless networks trends and challengesrdquo IEEE Com-puter Communications vol 11 no 5 pp 45ndash54 2005
[2] N Nasser A Hasswa and H Hassanein ldquoHandoffs in fourthgeneration heterogeneous networksrdquo IEEE CommunicationsMagazine vol 44 no 10 pp 96ndash103 2006
[3] M Ylianttila J Makeela and PMahonen ldquoVehicular telematicsover heterogeneous wireless networks a surveyrdquo ComputerCommunications vol 33 no 7 pp 775ndash793 2010
[4] X Liu L-G Jiang andCHe ldquoVertical handoff algorithmbasedon fuzzy logic in aid of pre-decision methodrdquo Acta ElectronicaSinica vol 35 no 10 pp 1989ndash1993 2007
[5] Y B Kang K Xu Y L Shen et al ldquoOptimal distributed verticalhandoff strategies in vehicular heterogeneous networksrdquo IEEEJournal on Communications supplement 1 pp 69ndash73 2009
[6] D Ma and M Ma ldquoA QoS-based vertical handoff scheme forinterworking of WLAN and WiMAXrdquo in Proceedings of theIEEEGlobal Telecommunications Conference (GLOBECOM rsquo09)pp 1ndash6 Honolulu Hawaii USA December 2009
[7] M Liu Z-C Li X-B Guo and K Zheng ldquoA movementtrend based self-adaptive vertical handoff algorithm and itsperformance evaluationrdquo Chinese Journal of Computers vol 31no 1 pp 112ndash119 2008
[8] Y Li M Li B Cao Y Wang and W Liu ldquoDynamic optimiza-tion of handover parameters adjustment for conflict avoidance
10 Chinese Journal of Engineering
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
in long term evolutionrdquo China Communications vol 10 no 1pp 56ndash71 2013
[9] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010
[10] W Tian Y Zhao Y Zhong M Xu and C Jing ldquoDynamic andintegrated load-balancing scheduling algorithm for cloud datacentersrdquo China Communications vol 8 no 6 pp 117ndash126 2011
[11] K Shafiee A Attar and V C M Leung ldquoOptimal distributedvertical handoff strategies in vehicular heterogeneous net-worksrdquo IEEE Journal on Selected Areas in Communications vol29 no 3 pp 534ndash544 2011
[12] J Zhang H C B Chan and V C M Leung ldquoA location-based vertical handoff decision algorithm for heterogeneousmobile networksrdquo in Proceedings of the IEEEGlobal Telecommu-nications Conference (GLOBECOM rsquo06) pp 1ndash5 San FranciscoCalif USA December 2006
[13] S Lee K Sriram K Kim Y H Kim and N Golmie ldquoVerticalhandoff decision algorithms for providing optimized perfor-mance in heterogeneous wireless networksrdquo IEEE Transactionson Vehicular Technology vol 58 no 2 pp 865ndash881 2009
[14] H Zhai X Chen and Y Fang ldquoHow well can the IEEE 80211wireless LAN support quality of servicerdquo IEEE Transactions onWireless Communications vol 4 no 6 pp 3084ndash3094 2005
[15] L-D Chou J-M Chen H-S Kao S-FWu andW Lai ldquoSeam-less streaming media for heterogeneous mobile networksrdquoMobile Networks and Applications vol 11 no 6 pp 873ndash8872006
[16] Y Sun F Yang and Z Shi ldquoDistributed network managementsystem with load balancingrdquo IEEE Journal on Communicationsvol 30 no 3 pp 34ndash41 2009
[17] Z-Y Feng P Zhang and Y-J Zhang ldquoA joint load control algo-rithm with terminal selection for the reconfigurable systemsrdquoJournal of Electronics and Information Technology vol 31 no 4pp 893ndash896 2009
[18] X Yan N Mani and Y A Sekercioglu ldquoA traveling distanceprediction based method to minimize unnecessary handoversfrom cellular networks to WLANsrdquo IEEE CommunicationsLetters vol 12 no 1 pp 14ndash16 2008
