Research ArticleAn Energy Efficient Data Gathering in Dense Mobile WirelessSensor Networks
R. Velmani1 and B. Kaarthick2
1 CSE Department, Anna University, Regional Centre, Coimbatore, Tamilnadu 641047, India2 ECE Department, Coimbatore Institute of Engineering and Technology, Coimbatore, Tamilnadu 641109, India
Correspondence should be addressed to R. Velmani; [email protected]
Received 6 February 2014; Accepted 20 March 2014; Published 16 April 2014
Academic Editors: G. Mazzini, A. Song, and Y. Yu
Copyright © 2014 R. Velmani and B. Kaarthick.This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.
Amidst of the growing impact of wireless sensor networks (WSNs) on real world applications, numerous schemes have beenproposed for collecting data on multipath routing, tree, clustering, and cluster tree. Effectiveness of WSNs only depends on thedata collection schemes. Existing methods cannot provide a guaranteed reliable network about mobility, traffic, and end-to-endconnection, respectively. To mitigate such kind of problems, a simple and effective scheme is proposed, which is named as clusterindependent data collection tree (CIDT). After the cluster head election and cluster formation, CIDT constructs a data collectiontree (DCT) based on the cluster head location. In DCT, data collection node (DCN) does not participate in sensing, which is simplycollecting the data packet from the cluster head and delivering it into sink. CIDT minimizes the energy exploitation, end-to-enddelay and traffic of cluster head due to transfer of datawithDCT.CIDTprovides less complexity involved in creating a tree structure,which maintains the energy consumption of cluster head that helps to reduce the frequent cluster formation and maintain a clusterfor considerable amount of time. The simulation results show that CIDT provides better QoS in terms of energy consumption,throughput, end-to-end delay, and network lifetime for mobility-based WSNs.
1. Introduction
WSNs have recently come into prominence because they holdpotential to revolutionize many segments of our economicallife, environmental monitoring, health care applications,infrastructure protection, context-aware computing, and bat-tlefield awareness [1]. The strength of WSNs lies in theirflexibility, energy consumption, mobility, and scalability. Thenumber of sensors capability and their organized fashionmade wireless sensor communication first option to utilizethem in remote or hazardous environments. The ultimategoal of such WSNs is often to deliver the sensing data fromsensor nodes to sink node and then conduct further analysisat the sink node [2]. To perform such tasks effectively, severalnetwork routing protocols have been proposed mainly fordata collection.
Topology management plays a vital role in minimizingvarious constraints such as limited energy, computationalresource crisis, latency, and quality of communication. Now,
the transmission distance between the sensor nodes isresponsible for energy consumption. Power loss is alwaysdirectly proportional to the distance 𝑃loss = 𝑑
𝜌 , where 𝑑 isthe distance between sensor nodes, 𝜌 is the environmentalfading factor, 𝜌 = 2 for free space fading, and 𝜌 = 4
for multipath fading [3]. The topology inherently definesthe types of routing path as broadcast or unicast and itdetermines the size, type of packets, and other overheads.Choosing a right topology helps to reduce the communica-tion overhead and energy conservation. An efficient topologyensures that neighbors are at a minimal distance and reducesthe probability of a packet being lost between sensor nodes.An efficient topology management may diminish the longrange communication within a network, communicationfailure and improves the network lifetime.
In addition, topologies in WSNs define the dimensionof the sensor node group and managing the addition ofnew members as well as dealing with members who left thegroup. By considering such aspects, the topologymay provide
Hindawi Publishing CorporationISRN Sensor NetworksVolume 2014, Article ID 518268, 10 pageshttp://dx.doi.org/10.1155/2014/518268
2 ISRN Sensor Networks
Sensor nodes
(a)
Sink
Sensor nodes
Cluster heads
(b)
Figure 1: (a) Flat topology, (b) cluster based topology.
an efficient data collection with low energy utilization andform superior WSN. The existing WSNs topologies are flat,tree, cluster, cluster tree, and chain. Based on the nature ofnetwork, different kinds of topologies are followed to gainthemaximumdata collection efficiency.This paper deals withan existing data collection topology and the proposed logicaltopology called DCT. It overcomes the existing limitationssuch as network lifetime andminimizes the energy consump-tion with effective data collection [4].
2. Related Work
Network topology determines the overall efficiency of theWSNs. Based on the data gathering and disseminationapplications, various types of logical topologies are definedinto (i) flat topology, (ii) cluster based topology, (iii) chainbased topology, (iv) tree based topology, (v) cluster treetopology.
2.1. Flat/Unstructured Topology (FT). FT/UT is a very simplemethod to collect the data for the sink [5]. FT is used in thecase of no topology or the absence of any defined topologyshown in Figure 1(a) (i.e., flooding and gossiping). Here,each sensor node plays an equal role to form a network.FT construction is a costly operation. It does not botherabout the energy constraints which lead to the implosionand overlapping problems [6]. For example, sensor protocolsfor information via negotiation (SPIN), directed diffusion,energy-aware routing, rumor routing, gradient based routing(GBR), constrained anisotropic diffusion routing (CADR),and cougar and active query forwarding in sensor networks(ACQUIRE) [7–14].
2.2. Chain Topology (CT). The CT constructs a transmissionchain to connect the deployed sensor nodes. A node isselected in the chain to act as leader of chain. All thesensor nodes can communicate with each other along thechain. Excessive delay for distant nodes on the chain is themain demerits of this topology (i.e., increasing the length
of the chain causes excessive delay where the leaf nodescollected data to reach the leader). When the sensor nodeshave high mobility, it leads to the link break problemsand affects the network performance. For example, Greedyalgorithm, minimum transmission energy (MTE), powerefficient gathering in sensor information systems (PEGASIS),chain oriented sensor network (COSEN), and chain routingwith even energy consumption (CREEC) [15–17].
2.3. Cluster Based Topology (CBT). CBT has been widelyused in WSNs for data gathering, data dissemination, andtarget tracking. Clustering is a proficient method for spe-cific applications, which requires scalability to hundredsor thousands of nodes (i.e., widely used in dense WSNs)shown in Figure 1(b). Scalability in this context implies theneed for load balancing, proficient resource exploitation, anddata aggregation. In clustering, cluster head election is animportant task. Here, the cluster head election is done byvarious methods like distributed (i.e., cluster head can beelected with probabilistic, residual energy, random method,and election phase) and centralized (i.e., cluster head havebeen assigned with nonprobabilistic methods by sink or basestation) election [18, 19].
After the cluster head election, all the cluster heads for-ward the data to the base station with direct hopping (clusterhead directly connected with base station) or multihopping(cluster head to cluster head communication) techniques.Formobility-based environments, frequent changes of clusterhead andmultihop techniques cannot offer a guaranteed datatransmission rate. It diminishes the performance of the entirenetwork. For example, low energy adaptive clustering hierar-chy (LEACH), hybrid energy efficient distributed clustering(HEED), base-station controlled dynamic clustering proto-col (BCDCP), concentric clustering scheme (CCS), energyaware routing protocol (EAR), hierarchical geographic mul-ticast routing (HGMR), cluster head gateway switch routing(CGSR), and mobility-based clustering protocol (MBC) [20–26].
2.4. Tree Based Topology (TBT). In TBT, all the deployedsensor nodes construct a logical tree. Generally, TBT workswith DFS (depth first search) or BFS (breadth first search)method [2]. Here, the entire data packet passes from leaf nodeto the parent nodes. Likewise, data flow from all sensor nodesto the sink is carried out. Constructing a logical tree avoidspacket flooding. It uses unicast instead of broadcast, as theflooding is not necessary for data communication.Therefore,tree topology consumes less power than flat topology. Whencompared with a few basic clustering protocol, tree topologyproves to be much more effective on energy utilization [27].Tree formation for the whole network is a time consumingand costly operation. It cannot tolerate with node failuresand power consumption is uneven across the network. Foravoiding the interference problem, different access methodsare chosen. Otherwise, it causes delay in sending the datapacket from leaf nodes to root node, for example, minimumspanning tree (MST), tree based data collection scheme(TBDCS), and efficient convergecast tree (ECT) [28–30].
ISRN Sensor Networks 3
Cluster head
Designated device
Figure 2: Cluster tree topology.
2.5. Cluster Tree Topology (CTT). CTT contains cluster andtree topology formation process shown in Figure 2. Thenetwork design starts with a special node called designateddevice (DD). It acts as a cluster head with greater trans-mission power and receiver sensitivity. The beacon signalcontains NetID, CID and NID nodes are added to the DD.Whenever, the node receives a beacon from a neighbornode, which sends a CONNECT REQUEST to DD. TheDD acknowledges to the corresponding node with ConnectResponse and the cluster tree formed. Here, the creation ofsuch topology with node id is a tedious process. Then, thespecial nodes (DD) should be initiated to make cluster tree[31], for example, ZigBee, 6LoWPAN. The main objective ofcluster tree is increasing the network capacity,minimizing theenergy consumption and end-to-end delay. But, the effective-ness of cluster tree is based on the network parameters likescalability, data rate, cluster dimension (number of clustersand cluster members for each cluster), tree intensity (numberof layers), RSS (received signal strength) and mobility (nodeposition, velocity, and direction), for example, cluster treedata gathering algorithm (CTDGA) [32].
2.6. Mobility Model. The mobility model is designed todescribe the location, velocity, and direction change over atime of mobile sensor nodes. The random waypoint model(RWM) is used in mobility management schemes (e.g., adhoc networks and sensor networks) [33]. The node travelsfrom a starting coordinate to a random ending coordinatewith a randomly generated constant velocity. The velocity ispicked from [0, 𝑉max] interval. When a sensor node reachesthe destination point, the node waits for a 𝑇pause time earlierthan arriving at the next destination [34].
3. Problem Statement
In flat topology, all the sensor nodes directly communicatewith the sink or simply forward the data packets to theneighbor nodes. Whenever, the sensor node wants to com-municate with a sink, the existing methods have limitationsuch as delay, data redundancy, and large amount of energyexploitation. Since, it is using flooding, gossiping, direct com-munication, and so forth. The cluster based data collectionsuggests better performance with cluster heads. Conversely,the data dissemination from cluster head to cluster head or
DCN
DCT
Sink Cluster head
Figure 3: CIDT structure.
sink (cluster head to sink communication must be eitherdirect hop or multihop communication) involves reliablestable links, which causes more energy consumption. Formobility-based environments, frequent cluster changes of thesensor node lead to link failure which causes diminishing thenetwork lifetime.
CT provides better performance than flat and clustertopology. However, it increases the data collection time thanCBT. Since, it must follow the chain route to reach sink, theentire network dies slowly due to the even energy utilizationof overall WSN. TBT can save more energy than CBT. Itincludes several time stamps in order to collect data from leafto root node. In mobility environment, it leads to link failure,packet drop, and delayed transmissions.
CTT offers enhanced performance than FT, CT, CBT,and TBT. The cluster head (DD) selection, maintaining thecluster with stable links for mobile sensor nodes is a costyoperation. The above topologies are not feasible and mendedadapt to mobile sensor ambiance. To overcome the existinglimitations in the above FT, CT, CBT, TBT, and CTT, wepropose a novel logical topology for data collection, namely,cluster independent data collection tree (CIDT).
Figure 3 shows the simple outline of our proposed schemenamed into CIDT structure. It is a unique nature of logicalscheme, which helps to improve the network lifetime andeffective data collection, thereby increasing network lifetimewith minimum delay. CIDT is a best hybrid scheme (whichutilizes cluster and tree topology) suitable for dense wire-less sensor networks than any other logical topology. Onmobility-based environments, it provides better performancethan other methods.
4. CIDT (Cluster Independent DataCollection Tree)
The CIDT consists of setup phase and steady state phase. Insetup phase, cluster formation and tree construction is initi-ated to identify the optimal path between clustermember andsink. It is denoted in intracluster and DCT communication.DCT construction for single cluster is shown in Figure 4.
Now, the cluster head is responsible for the data collectionfrom cluster members and cluster maintenance operations.At first, all the sensor nodes elect ahead to the cluster headand form a cluster. Thereafter, tree formation is initiated,
4 ISRN Sensor Networks
DCN
Cluster head
Cluster member
DCT
Tree link tonext cluster
Figure 4: CIDT in single cluster.
which connects the cluster head and sink. Here, the clusterformation and DCT construction is based on the thresholdvalue, connection time, and RSS. After the setup phasecompletion, data transmission is initiated in steady statephase. Here, all the cluster members send ahead the datapackets to sink based on the optimal path.
4.1. CIDT Tree Formation. For a large-scale WSN, numerousnumber of sensor nodes have been randomly deployed. Inthis case, the selection of DCN does not affect the datacollection of a corresponding cluster. It should have betterconnection timewith the nearestDCNnode and cluster head.The DCT formation is based on the location of cluster head,connection time between the cluster head and DCN. Afterthe cluster head election, BS or sink initiates ahead to theDCT formation process. Based on the location of cluster headand connection time, a few numbers of nodes are selected asDCN. Now, the DCN may act as a data collection node anddoes not participate in sensing. But, it does not belong to anycluster.
All the DCN collects the data from cluster head, whichaggregate with the corresponding cluster head and thenforward to the next DCN. The DCN selection algorithm isexecuted by sink in order to select the DCN to form anindependent tree structure. Figure 5 presents the flow chartof DCT construction.Algorithm 1 steps:
(a) Initialize the values.
(b) Choose a random TIN (temporary independentnode) from sink.
(c) In case, a very first one-hop distance node fromthe sink is CH. Then, skip to another one-hop NN(Neighbor Node).
(d) Select a one-hop NN as CNI (Current Node Identity)and assign the integrity as TIN.
(e) Now, the one-hop distance (NH, NN) CNI from thesink is considered into TIN and select a better TINamong them.
(f) Select a one-hop distance NN to the CH as PIN fromTIN.
Yes
No
Start
Deploy the sensor nodes (N)
(1) Elect the CH over an entire network(2) Initialize the count i, j, m, N,
NH = 1, HC = 1
Elect a one-hop distance SN from thesink as CNI to select DCN for DCT generation
If (CNI ≠ CH)
No
Yes
Stop
Select FSIN from random TIN or SIN and
Select TIN from CNI and increment NH, NN
If (NN == CH)
among the CM from appropriate cluster (CH ≠ NH)
Construct DCT with DCN
Randomly select one SN with HC → 1 as SIN
change the integrity of FSIN to DCN
Figure 5: Flow chart of DCT construction.
(g) In that case, step (c) is not able to find TIN.Then, skipto select a SIN from TIN, which is another one-hopdistance NN (Neighbor Node) from CH.
(h) Increment the value of NH and NN for next DCNselection.
(i) Change the integrity of SIN (Selected IndependentNode) to FSIN (Finalized and Selected IndependentNode).
(j) Choose a FSIN from random TIN.(k) Let the integrity of FSIN considered as DCN, which
is used to construct a DCT link between the sink andcluster head.
The above list represents the algorithm for DCT. Initiallysink starts with the one-hop neighbor sensor node to addthat particular node to act as a DCN in DCT.The parametersinclude HC = 1 (hop distance is used to select a one-hopneighbor node from sink to act as a current node identity(CNI)), NH (new hop distance is an additive value, whichdenotes the current distance of node (CNI) from sink andit is used to finalize the DCN selection). Then, the identifiednodes have been stored in temporary structure (TIN). In case,the one-hop distance neighbor node (NN) is found to be CH,one node from the cluster head with HC = 1 is identified asTIN. After finding the NH of the network, starting with thenearest node as cluster head, the node selection is finalized.Then selected nodes are utilized to form the DCT.
ISRN Sensor Networks 5
DCT (CNI, NH, HC, NN)int i, j,m, N, NH = 1, HC = 1 (a)for (i ← 1 to N) (b){ if (CNI == CH){
NH = NH++ (c)TIN ← NH (d)}
else{
if (CNI == NN && CNI = CH++ && TIN == CNI)for (j ← 0 tom) (e)
{
PIN ← TIN (f)} }
else{
SIN ← TIN (g)NH ← NH++NN ← NH++ (h)
}
if (PIN == TIN && SIN == TIN && SIN == PIN){
FSIN ← SIN (i)}
else{
FSIN ← Random (TIN) (j)}
DCN = FSINDCT ← DCN (k)
}
Algorithm 1: DCT Construction.
DCT is a hierarchical tree structure, which covers theentire WSNs. DCN collects the data from the cluster headsand delivers it to sink or BS. Selecting DCN with betterconnection time and best communication range reducesenergy consumption due to long range node to node datatransfer. While the sensor nodes are on high mobility, theselectedDCNkeeps the communicationwith the cluster headfor a longer time and there is no need to update the treestructure. In order to keep the lifetime of whole network inharmony DCN is also newly selected every time when thenew cluster heads are elected. New DCN selection also iscarried out by sink, which is based on the mobility of the newcluster head.
4.2. Intracluster Communication. Considering ambiguouslarge-scale WSNs, sensor nodes have been densely deployedover the region. During the setup phase, the beacon signal isused to identify the sensor nodes location and position. Oncethe nearby nodes are identified, randomalgorithmor electionalgorithm is used to elect the cluster head. After the clusterhead selection, the next phase DCT formation is initiated.
In the proposed method, the threshold value 𝑈𝑛
𝑎𝑏has
been calculated in (1) by adding the flag value with themultiplication of factors such as the total number of neighbornodes, residual energy, current speed, and current coveragedistance of the sensor node, where 𝐹𝐶 is the flag (set 𝐹𝐶 =
1 for previous round cluster head and 𝐹𝐶 = 0 for sensornode having a chance to act as current round cluster headbased on 𝑈
𝑛
𝑎𝑏), 𝐸𝑛-current is the current sensor node energy,
𝑉𝑛-current is the current speed of the sensor node, 𝐸max isthe initial energy, and 𝑉max is the maximum speed of thesensor node. In order to avoid the election of high mobilitynode as cluster head,((𝑉max − 𝑉𝑛-current)/(𝑉max + 𝑉𝑛-current))instead of (𝑉max/𝑉𝑛-current) may be considered. The expectednumber of sensor nodes in each cluster is𝑚 = 𝑁𝑆/𝐶𝐻. Thosenodes having maximum residual energy, maximum numberof cluster members, and maximum connection time can beelected as cluster head:
𝑈𝑛
𝑎𝑏= (𝐹𝐶 + (
𝑁𝑏-current𝑁max − 𝑁𝑏-current
×𝐸max − 𝐸𝑛-current
𝐸max
×𝑉max − 𝑉𝑛-current𝑉max + 𝑉𝑛-current
)) .
(1)
It is visualized that the 2D network position of the clusterhead b and sensor node 𝑎 at time 𝑡 is characterized in thefollowing:
𝑋𝑏 = 𝑥𝑏 + V𝑏 ∗ cos 𝜃𝑏𝑡; 𝑌𝑏 = 𝑦𝑏 + V𝑏 ∗ sin 𝜃𝑏𝑡
𝑋𝑎 = 𝑥𝑎 + V𝑎 ∗ cos 𝜃𝑎𝑡; 𝑌𝑎 = 𝑦𝑎 + V𝑎 ∗ sin 𝜃𝑎𝑡,
(2)
6 ISRN Sensor Networks
where (𝑥, 𝑦) is the primary node location, V is the speed, 𝜃 isthemoving path angle between (𝑥, 𝑦), and 𝑡 is the connectiontime. Then, the subscript (𝑎, 𝑏) corresponds to sensor node 𝑎
and cluster head 𝑏, respectively. Let the 𝑅𝑡
𝑎𝑏be denoted as
𝑅𝑡
𝑎𝑏≥ ((𝑋𝑏 − 𝑋𝑎)
2+ ( 𝑌𝑏 − 𝑌𝑎)
2)1/2
. (3)
At time 𝑡 = 0, each sensor node receives an advertisementmessage from any one of the cluster heads. Hence, the above2D network equation (??) is considered and simplified into
𝑅𝑡
𝑎𝑏≥ ((𝑥𝑏 − 𝑥𝑎)
2+ ( 𝑦𝑏 − 𝑦𝑎)
2)1/2
if 𝑡 = 0. (4)
Now, Δ𝑅𝑡,𝑡+𝑠
𝑎𝑏is the difference between 𝑅
𝑡
𝑎𝑏and 𝑅
𝑡+𝑠
𝑎𝑏at time
instance 𝑡 and 𝑡 + 𝑠. Let Δ𝑅𝑡,𝑡+𝑠
𝑎𝑏be found using (4):
Δ𝑅𝑡,𝑡+𝑠
𝑎𝑏= 𝑅𝑡
𝑎𝑏− 𝑅𝑡+𝑠
𝑎𝑏if (𝑛, 𝑠) ∈ 𝑡, 𝑠 = 0, 1, 2. . . . 𝑛. (5)
However, forΔ𝑅𝑡,𝑡+𝑠
𝑎𝑏= 0, there is no sensor nodes onmobility
within a cluster. Δ𝑅𝑡,𝑡+𝑠
𝑎𝑏is the negative value for sensor nodes
in a cluster moving away from the cluster head; Δ𝑅𝑡,𝑡+𝑠
𝑎𝑏is
the positive value cluster head and cluster member movingtowards to each other. Now, the RSS (received signal strength)can be calculated at any time instance 𝑡 and 𝑡 + 𝑛 in
RSS𝑡𝑎𝑏
= RSS𝑡𝑎𝑏-current − RSS𝑎𝑏-min
RSS𝑡+𝑠𝑎𝑏
= RSS𝑡+𝑠𝑎𝑏-current − RSS𝑎𝑏-min ∀𝑠 ∈ 𝑡,
(6)
whereas RSS𝑎𝑏-min is the minimum required threshold valueandRSS𝑡+𝑠
𝑎𝑏-current is the current threshold value at time instance𝑡. If RSS is a positive value, the cluster members join inan appropriate cluster and communicate with correspondingcluster head. In this case, ΔRSS𝑡,𝑡+𝑠
𝑎𝑏is the difference between
RSS𝑡𝑎𝑏and RSS𝑡+𝑠
𝑎𝑏, which can be found from
ΔRSS𝑡,𝑡+𝑠𝑎𝑏
= RSS𝑡𝑎𝑏
− RSS𝑡+𝑠𝑎𝑏
, (7)
wherever, ΔRSS𝑡,𝑡+𝑠𝑎𝑏
≤ 0, Cluster member move away fromthe current position of the cluster head. ΔRSS𝑡,𝑡+𝑠
𝑎𝑏≥ 0, both
cluster member and cluster head move towards each otherfrom their current position. Gn
ab is the value assigned tosensor node a for each round, which indicates its robustnessfor connection with cluster head b. The dimensionless value𝛿G, 𝜁G, 𝜂G, and 𝜅CT is a linear combination with constantcoefficients between 0 and 1. The coefficients represent theconsequence of each factor and are denoted as follows:
𝛿𝐺 + 𝜁𝐺 + 𝜂𝐺 + 𝜅𝐶𝑇 = 1. (8)
Therefore, (8) can be originated into 𝐺𝑛
𝑎𝑏and in (9) repre-
sented as
𝐺𝑛
𝑎𝑏= (𝛿𝑇 × (
𝐸max − 𝐸𝑏-current𝐸𝑏-current × 𝑁𝑏-current
))
+ (𝜁𝑇 × (1 −RSS𝑎𝑏-min
RSS𝑎𝑏-current))
+ (𝜂𝑇 × (𝑑𝑏𝑎 − 𝑅
𝑡
𝑎𝑏
𝑅𝑡
𝑎𝑏
)) + (𝜅𝐶𝑇 × (Δ𝑡𝑏
𝑎𝑏
𝑡𝑓
𝑐
)) ,
(9)
where 𝐸max is the initial energy, 𝐸𝑏-current is the cluster headcurrent energy, 𝑁𝑏-current is the number of current clustermembers for cluster head 𝑏, RSS𝑎𝑏-min is the minimumrequired RSS from 𝑎 and 𝑏, RSS𝑎𝑏-current is the current RSSbetween a and 𝑏, 𝑅
𝑡
𝑎𝑏is the distance between a and b at
any time instance 𝑡, 𝑑𝑏𝑎 is the maximum coverage distancebetween b and a, Δ𝑡
𝑏
𝑎𝑏is the estimated connection time for a
begins its transmission to b, and 𝑡𝑓
𝑐is the current duration of
the data frame for b.
4.3. DCT Communication. After the intracluster communi-cation phase, DCT formation phase is initiated. It is basedon the threshold value, connection time, and network traffic.DCT makes a communication link between the cluster headand sink. Let it be visualized that the 2D network position ofthe cluster head b and DCN e or h at time t is
𝑋𝑏 = 𝑥𝑏 + V𝑏 ∗ 𝑐𝑜𝑠𝜃𝑏𝑡; 𝑌𝑏 = 𝑦𝑏 + V𝑏 ∗ sin 𝜃𝑏𝑡
𝑋𝑒 = 𝑥𝑒 + V𝑒 ∗ cos 𝜃𝑒𝑡; 𝑌𝑒 = 𝑦𝑒 + V𝑒 ∗ sin 𝜃𝑒𝑡
𝑋ℎ = 𝑥ℎ + Vℎ ∗ cos 𝜃ℎ𝑡; 𝑌ℎ = 𝑦ℎ + Vℎ ∗ sin 𝜃ℎ𝑡.
(10)
On each round, the distance between cluster head b to DCN eand h has been calculated from (10). Let the distance 𝑅
𝑡
𝑏𝑒and
𝑅𝑡
𝑒ℎbe denoted in
𝑅𝑡
𝑏𝑒≥ ((𝑋𝑏 − 𝑋𝑒)
2+ ( 𝑌𝑏 − 𝑌𝑒)
2)1/2
.
𝑅𝑡
𝑒ℎ≥ ((𝑋𝑒 − 𝑋ℎ)
2+ ( 𝑌𝑒 − 𝑌ℎ)
2)1/2
, 0 ≤ 𝑡 ≤ 𝑛
(11)
Let 𝑡 = 0; the distance 𝑅𝑡
𝑏𝑒and 𝑅
𝑡
𝑒ℎin (11) can be considered
as
𝑅𝑡
𝑏𝑒≥ ((𝑥𝑏 − 𝑥𝑒)
2+ (𝑦𝑏 − 𝑦𝑒)
2)1/2
𝑅𝑡
𝑒ℎ≥ ((𝑥𝑒 − 𝑥ℎ)
2+ (𝑦𝑒 − 𝑦ℎ)
2)1/2
, 𝑡 = 0, ∀𝑛 ∈ 𝑡.
(12)
For any cluster head toDCNorDCN toDCNor sink toDCNcommunication, the threshold value 𝑈
𝑝
𝑢V has been calculatedin (13) by adding total number ofDCNwithmultiplied factorssuch as the residual energy, and current speed between clusterhead to DCN or DCN to DCN. Let 𝑢 be considered insteadof 𝑏 (it represents that cluster head or sink or DCN), V as asubstitute of 𝑒 and ℎ (it signifies that DCN):
𝑈𝑝
𝑢V = 𝐻V + (𝐸V-max − 𝐸V-current
𝐸V-max×
𝑉V-max− 𝑉V-current𝑉V-max + 𝑉V-current
) , (13)
where𝐻V is the count for DCN,𝐸V-current is the current clusterhead energy,𝑉V-current is the current speed of the cluster head,𝐸V-max is the initial energy, and 𝑉V-max is the maximum speedof cluster head. Let Δ𝑅
𝑡,𝑡+𝑛
𝑢V be the diversity with Rtuv and R
t+nuv .
At the time instance t and t + n, ΔRt, t+nuv is represented in
Δ𝑅𝑡,𝑡+𝑛
𝑢V = 𝑅𝑡
𝑢V − 𝑅𝑡+𝑛
𝑢V , 𝑡 = 0, 1, 2 . . . 𝑛, ∀𝑛 ∈ 𝑡. (14)
However, Δ𝑅𝑡,𝑡+𝑛
𝑢V = 0, both nodes (cluster head and DCNor any two DCN) not in mobility, which is separated in even
ISRN Sensor Networks 7
distance. Δ𝑅𝑡,𝑡+𝑛
𝑢V is the negative value for both nodes movingaway; Δ𝑅
𝑡,𝑡+𝑛
𝑢V is the positive value for both nodes movingtowards each other. Consequently, the RSS (received signalstrength) between any two nodes, at the time instance 𝑡 and𝑡 + 𝑛 is calculated in
RSS𝑡𝑢V = RSS𝑡
𝑢V-current − RSS𝑢V-min
RSS𝑡+𝑛𝑢V = RSS𝑡+𝑛
𝑢V-current − RSS𝑢V-min ∀𝑛 ∈ 𝑡,
(15)
where RSS𝑢V-min is the minimum required threshold valueand RSS𝑡+𝑛
𝑢V-current is the current threshold value. If RSS𝑡+𝑛𝑢V
is a positive value, then the cluster head or DCN has alikelihood to join with nearest DCN, which can establish thecommunication with corresponding nodes. ΔRSS𝑡,𝑡+𝑛
𝑢V can befound using (15) as follows:
ΔRSS𝑡,𝑡+𝑛𝑢V = RSS𝑡
𝑢V − RSS𝑡+𝑛𝑢V , ∀𝑛 ∈ 𝑡 (16)
wherever ΔRSS𝑡,𝑡+𝑛𝑢V ≤ 0, both nodes moving away from their
current position. ΔRSS𝑡,𝑡+𝑛𝑢V ≥ 0, both nodes moving towards
each other.Thedimensionless value𝛿𝐶𝑇, 𝜁𝐶𝑇, 𝜂𝐶𝑇, and 𝜅𝐶𝑇 is alinear combination with constant coefficients between 0 and1. The coefficients represent the consequence of each factorand are denoted as
𝛿𝐶𝑇 + 𝜁𝐶𝑇 + 𝜂𝐶𝑇 + 𝜅𝐶𝑇 = 1. (17)
𝐺𝑝
𝑢V is the value assigned to all 𝑢 on each round, whichindicates its heftiness for connection with V:
𝐺𝑝
𝑢V = (𝛿𝐶𝑇 × (𝐸V-max − 𝐸V-current
𝐸V-current × 𝑁V-current))
+ (𝜁𝐶𝑇 × (1 −RSS𝑢V-minRSS𝑢V-current
))
+ (𝜂𝐶𝑇 × (𝑑𝑢V − 𝑅
𝑡
𝑢V
𝑅𝑡𝑢V
))
+ (𝜅𝐶𝑇 × (Δ𝑡
V𝑢V
𝑡𝑓
𝐶𝑇(Ψ)
)) ,
(18)
where 𝐸V-max is the initial energy of v, 𝐸V-current is the currentenergy of V, 𝑁V-current is the total number of cluster head orDCN connected with V, RSS𝑢V-min is the minimum requiredRSS to make a connection from 𝑢 and V, RSS𝑢V-current is thecurrent RSS to establish a connection between u and V, 𝑑𝑢V isthe maximum coverage distance between 𝑢 and V, 𝑅𝑡
𝑢V is thedistance between 𝑢 and V at any time instance 𝑡, Δ𝑡
V𝑢V is the
estimated connection time for 𝑢 begins its transmission to V,and 𝑡𝑓
𝐶𝑇(Ψ) is the current duration of the data frame for V.
4.4. Frame Duration. Let us consider the number of currentcluster Members Mc and the number of expected clustermembers 𝑀𝑒 can be derived from the following equation:
𝑀𝑐 = 𝑀𝑒 − (𝑀𝑑 + 𝑀𝑠)
𝑀𝑒 =𝑁𝑐 − 𝐶𝐻 − 𝐶𝑇
𝐶𝐻
; 𝑁𝑐 = 𝑁𝑡 − 𝑁𝑑,(19)
where𝑀𝑐 is the current cluster member from one cluster,𝑀𝑒is the expected number of cluster member,𝑀𝑑 is the number
100 200 300 400 500
85
90
95
100
Number of nodes
Pack
et d
eliv
ery
ratio
(%)
CIDTMBC
HEEDLEACH
Figure 6: Packet delivery ratio versus number of nodes.
100 200 300 400 500
0.2
0.4
0.6
0.8
1.0
Number of nodes
Thro
ughp
ut (M
bps)
CIDTMBC
HEEDLEACH
Figure 7: Throughput versus number of nodes.
of cluster member dead, 𝑀𝑠 is the total number of clustermembers on sleep state, 𝑁𝑐 is the total number of currentsensor nodes, 𝑁𝑡 is the total number of sensor nodes over anetwork, 𝑁𝑑 is the number of sensor nodes dead, and 𝐶𝐻 isthe cluster head. Now, the current duration of the data frame𝑡𝑓
𝑐from each cluster is denoted in
𝑡𝑓
𝑐=
𝐿𝑝
𝑅𝑏
× 𝑀𝑐 or 𝑡𝑓
𝑐=
𝐿𝑝
𝑅𝑏
× 𝑀𝑒, (20)
where 𝐿𝑝 is the data packet length and 𝑅𝑏 is the transmissionbit rate.
4.5. Steady State Phase. On steady state phase, each clustermember and the corresponding cluster head build intraclus-ter communication with each other. Initially, all the clustermembers send the sensed data to cluster head in an allocatedTDMA time slot. Thereafter, the cluster head aggregatesthe received data, and then forward the data packet to theDCN. Again, DCN aggregates the data packet from its clusterhead and then forward to the sink with DCT. In DCTcommunication, direct sequence spread spectrum techniquescan be used to transfer the data packets from the clusterhead to DCN and sink. DCT discovers an optimal pathbetween the cluster head and the sink based on the distance,connection time, threshold value, and residual energy. Based
8 ISRN Sensor Networks
50
100
150
200
250
Tota
l ene
rgy
(mJ)
100 200 300 400 500
Number of nodes
CIDTMBC
HEEDLEACH
Figure 8: Total energy versus number of nodes.
100 150 200 250 300 350 400 450 500
2
4
6
8
10
Del
ay (m
s)
Number of nodes
CIDTMBC
HEEDLEACH
Figure 9: Delay versus number of nodes.
on the optimal path, the entire cluster head forwards thedata packets to the nearest DCN. Now, the DCT becomesresponsible for forwarding an entire data from the clusterhead to sink.
5. Results and Discussion
In this section, the simulation results are used to evaluate theperformance of the proposed protocol under various param-eter settings. The network simulator was used to carry outa performance study of CIDT to compare with LEACH andMBC. Considering 500 nodes of WSNs, all the nodes wererandomly deployed in a square region of 1000 × 1000m2, thesize of data packet is 512 bytes, the transmission range withinthe cluster 40m, the transmission range between the cluster80 to 120m, the sensing range is 20m, and the base stationis located in (𝑥 = 500, 𝑦 = 1050). Further communicationenergy parameters can be set as 𝐸elec = 50 nJ/bit/m2 and𝐸amp = 0.0013 pJ/bit/m4. Then, the energy required for dataaggregation is set into 𝐸DA = 50 nJ/bit/signal.
Based on CIDT, the network performance was simulatedin terms of the packet delivery ratio (PDR), throughput, delay,total energy, and speed. Figures 6, 7, 8, and 9 illustrate therelationship between the number of deployed nodes and theperformance of the network (PDR, throughput, total energyconsumption, and delay). It is worth noting that LEACH,
5 10 15 20 25 30
75
80
85
90
95
100
Speed (m/s)
Pack
et d
eliv
ery
ratio
(%)
CIDTMBC
HEEDLEACH
Figure 10: Packet delivery ratio versus speed.
5
10
15
20
25
30
Del
ay (m
s)
5 10 15 20 25 30
Speed (m/s)
CIDTMBC
HEEDLEACH
Figure 11: Delay versus speed.
HEED, and MBC fail to prolong the PDR, throughput,total energy consumption, and delay as the number of nodeincreases. However, CIDTmakes better performance linearlyeven the number of sensor node increases over the network.In large-scale mobility-based WSNs, unreliable links maycause the packet loss and retransmissions. In that case, itmay increase the energy consumption of sensor nodes. Inaddition, it may reduce the PDR and throughput. Although,CIDT can provide stable links and guarantee the balancedenergy conservation over the network. Therefore, it can beconclude that CIDT protocol has been mended adapting tothe high mobility environment.
Figures 10 and 11 show that CIDT has superior perfor-mance when compared to MBC, HEED, and LEACH inmobile sensor ambience. In the simulation results, it can beenunciated that CIDT protocol has provided stable links andmended adapting to the high mobility environment. On highmobility environment, CIDTmakes better PDR and less end-to-end delay.
Finally, it can be concluded that the proposed CIDTprotocol can save the sensor nodes residual energy, extendthe network lifetime and network reliability. It is mendedadapting to the high mobility environment with better com-munication quality.
ISRN Sensor Networks 9
6. Conclusions
With the growing impact of WSNs on real time civil andmilitary applications, numerous sensor nodes are requiredto monitor the large-scale areas. Cluster tree is a proficientmethod to construct suspicious network management archi-tecture. The ultimate goal is to exploit the network lifetime,residual energy, throughput, PDR, and stable link for mobilesensor nodes. In this paper, CIDT (cluster independentdata collection tree) proposed for mobility-based WSNs,each cluster member chooses the cluster head with betterconnection time, and RSS. Then, forward the data packetsto the corresponding cluster head in an allocated time slot.Consequently, the sink or DCN select the one-hop neighborDCN or cluster head with the maximum of threshold value,RSS, connection time, and less network traffic. From thesimulation results, it is evident that CIDT provides morestable links, throughput, PDR with a reduction of networktraffic and a condensed sum of energy utilization thanLEACH, HEED, and MBC.
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper.
References
[1] F. Zhao and L. J. Guibas, Wireless Sensor Networks : AnInformation Processing Approach, Elsevier, 2004.
[2] S. Chen, M. Huang, S. Tang, and Y. Wang, “Capacity ofdata collection in arbitrary wireless sensor networks,” IEEETransactions on Parallel and Distributed Systems, vol. 23, no. 1,pp. 52–60, 2012.
[3] W. B. Heinzelman, A. P. Chandrakasan, and H. Balakrish-nan, “An application-specific protocol architecture for wirelessmicrosensor networks,” IEEE Transactions onWireless Commu-nications, vol. 1, no. 4, pp. 660–670, 2002.
[4] Q. Mamun, “A qualitative comparison of different logicaltopologies for wireless sensor networks,” Sensors, vol. 12, no. 11,pp. 14887–14913, 2012.
[5] K.-W. Fan, S. Liu, and P. Sinha, “Structure-free data aggregationin sensor networks,” IEEE Transactions on Mobile Computing,vol. 6, no. 8, pp. 929–942, 2007.
[6] K. Akkaya and M. Younis, “A survey on routing protocols forwireless sensor networks,” Ad Hoc Networks, vol. 3, no. 3, pp.325–349, 2005.
[7] W. Heinzelman, J. Kulik, and H. Balakrishnan, “Adaptive proto-cols for information dissemination inwireless sensor networks,”in Proceedings of the 5th Annual ACM/IEEE InternationalConference on Mobile Computing and Networking (MobiCom’99), Seattle, Wash, USA, 1999.
[8] C. Intanagonwiwat, R. Govindan, and D. Estrin, “Directeddiffusion: a scalable and robust communication paradigm forsensor networks,” in Proceedings of the 6th Annual InternationalConference on Mobile Computing and Networking (MOBICOM’00), pp. 56–67, Boston, Mass, USA, August 2000.
[9] E. Astier, A. Hafid, and A. Benslimane, “Energy and mobilityaware clustering technique for multicast routing protocols in
wireless Ad hoc networks,” in Proceedings of the IEEE Wire-less Communications and Networking Conference (WCNC ’09),Orlando, Fla, USA, April 2009.
[10] D. Braginsky and D. Estrin, “Rumor routing algorithm forsensor networks,” in Proceedings of the 1st ACM InternationalWorkshop onWireless Sensor Networks and Applications (WSNA’02), pp. 22–31, Atlanta, Ga, USA, September 2002.
[11] C. Schurgers and M. B. Srivastava, “Energy efficient routingin wireless sensor networks,” in Proceedings of the MilcomCommunications for Network-Centric Operations: Creating theInformation Force, pp. 357–361, McLean, Va, USA, October2001.
[12] M. Chu, H. Haussecker, and F. Zhao, “Scalable information-driven sensor querying and routing for ad hoc heterogeneoussensor networks,”The International Journal of High PerformanceComputing Applications, vol. 16, no. 3, pp. 293–313, 2002.
[13] Y. Yao and J. Gehrke, “The cougar approach to in-network queryprocessing in sensor networks,” SIGMOD Record, vol. 31, no. 3,pp. 9–18, 2002.
[14] N. Sadagopan, B. Krishnamachari, and A. Helmy, “TheACQUIRE mechanism for efficient querying in sensor net-works,” in Proceedings of the 1st International Workshop onSensor Network Protocol and Applications, Anchorage, Alaska,USA, 2003.
[15] S. Lindsey and C. S. Raghavendra, “PEGASIS: power-efficientgathering in sensor information systems,” in Proceedings of theIEEEAerospace Conference Proceedings, vol. 3, pp. 1125–1130, BigSky, Mont, USA, 2002.
[16] N. Tabassum, Q. E. K. Mamun, and Y. Urano, “COSEN: achain oriented sensor network for efficient data collection,” inProceedings of the 3rd International Conference on InformationTechnology: New Generations (ITNG ’06), pp. 262–267, April2006.
[17] J. Shin and C. Suh, “CREEC: chain routing with even energyconsumption,” Journal of Communications and Networks, vol.13, no. 1, pp. 17–25, 2011.
[18] M. Ye, C. Li, G. Chen, and J. Wu, “EECS: an energy efficientclustering scheme in wireless sensor networks 10a.2,” in Pro-ceedings of the 24th IEEE International Performance, Computing,and Communications Conference (IPCCC ’05), pp. 535–540,Phoenix, Ariz, USA, April 2005.
[19] X. Liu, “A survey on clustering routing protocols in wirelesssensor networks,” Sensors, vol. 12, no. 8, pp. 11113–11153, 2012.
[20] O. Younis and S. Fahmy, “HEED: a hybrid, energy-efficient,distributed clustering approach for ad hoc sensor networks,”IEEE Transactions on Mobile Computing, vol. 3, no. 4, pp. 366–379, 2004.
[21] S. D. Muruganathan, D. C. F. Ma, R. I. Bhasin, and A. O.Fapojuwo, “A centralized energy-efficient routing protocol forwireless sensor networks,” IEEECommunicationsMagazine, vol.43, no. 3, pp. S8–S13, 2005.
[22] S.-M. Jung, Y.-J. Han, and T.-M. Chung, “The concentricclustering scheme for efficient energy consumption in thePEGASIS,” in Proceedings of the 9th International Conference onAdvanced Communication Technology (ICACT ’07), vol. 1, pp.260–265, Gangwon-Do, Korea, February 2007.
[23] M. Liu, J. Cao, G. Chen, andX.Wang, “An energy-aware routingprotocol in wireless sensor networks,” Sensors, vol. 9, no. 1, pp.445–462, 2009.
[24] D. Koutsonikolas, S. M. Das, Y. C. Hu, and I. Stojmenovic,“Hierarchical geographic multicast routing for wireless sensornetworks,”Wireless Networks, vol. 16, no. 2, pp. 449–466, 2010.
10 ISRN Sensor Networks
[25] H. Raza, P. Nandal, and S. Makker, “Selection of cluster-head using PSO in CGSR protocol,” in Proceedings of the 2ndInternational Conference on Methods and Models in ComputerScience, pp. 91–94, December 2010.
[26] S. Deng, J. Li, and L. Shen, “Mobility-based clustering protocolfor wireless sensor networks with mobile nodes,” IET WirelessSensor Systems, vol. 1, no. 1, pp. 39–47, 2011.
[27] Z. Zhang and F. Yu, “Performance analysis of cluster-based andtree-based routing protocols for wireless sensor networks,” inProceedings of the International Conference on CommunicationsandMobile Computing (CMC ’10), vol. 1, pp. 418–422, Shenzhen,China, April 2010.
[28] H. Li, H. Yu, and A. Liu, “A tree based data collection schemefor wireless sensor network,” in Proceedings of the InternationalConference on Networking, International Conference on Systemsand International Conference on Mobile Communications andLearning Technologies (ICN/ICONS/MCL ’06), p. 119, Morne,Mauritius, April 2006.
[29] D. Messina, M. Ortolani, and G. L. Re, “A network protocolto enhance robustness in tree-based WSNs using data aggre-gation,” in Proceedings of the IEEE Internatonal Conference onMobile Adhoc and Sensor Systems (MASS ’07), pp. 1–4, Pisa, Italy,October 2007.
[30] M. R. Saadat and G. Mirjalily, “Efficient convergecast tree fordata collection in wireless sensor networks,” in Proceedings ofthe 20th Iranian Conference on Electrical Engineering (ICEE ’12),pp. 1534–1539, Tehran, Iran, 2012.
[31] E. Callaway, “Cluster Tree Network—IEEE P802. 15 Wirelesspersonal AreaNetworks,” 2001, http://www.ieee802.org/15/pub/2001/May01/01189r0P802-15 TG4-Cluster-Tree-Network.pdf.
[32] J. Yang, B. Bai, and H. Li, “A cluster-tree based data gatheringalgorithm for wireless sensor networks,” in Proceedings of theInternational Conference on Automatic Control and ArtificialIntelligence (ACAI ’12), pp. 22–25, Xiamen, China, 2012.
[33] E. Hyytia, H. Koskinen, P. Lassila, A. Penttinen, and J. Virtamo,Random Waypoint Model in Wireless Networks, NetworkingLaboratory, Helsinki University of Technology, Helsinki, Fin-land, 2005.
[34] O. Demigha, W. K. Hidouci, and T. Ahmed, “On energyefficiency in collaborative target tracking in wireless sensor net-work : a review,” IEEE Communications Surveys and Tutorials,vol. 15, no. 3, pp. 1210–1222, 2013.
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporation http://www.hindawi.com
Journal ofEngineeringVolume 2014
Submit your manuscripts athttp://www.hindawi.com
VLSI Design
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation http://www.hindawi.com
Volume 2014
The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
DistributedSensor Networks
International Journal of