evidence-efficient multihop clustering routing scheme for...

Research ArticleEvidence-Efficient Multihop Clustering Routing Scheme forLarge-Scale Wireless Sensor Networks

Zhihua Li and Ping Xin

Department of Computer Science and Technology School of Internet of Things Engineering Jiangnan University WuxiJiangsu 214122 China

Correspondence should be addressed to Zhihua Li zhlijiangnaneducn

Received 22 August 2017 Accepted 30 October 2017 Published 19 November 2017

Academic Editor Paolo Barsocchi

Copyright copy 2017 Zhihua Li and Ping Xin 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

Energy consumption and transmission reliability are the most common issues in wireless sensor networks (WSNs) By studying thebroadcast nature of data transmission in WSNs the mechanism of guaranteeing reliable transmission is abstracted as propagationof responsibility and availabilityThe responsibility and availability represent the accumulated evidence of nodes to support reliabletransmission Based on the developedmechanism an evidence-efficient cluster head rotation strategy and algorithm are presentedFurthermore backbone construction algorithm is studied to generate the minimum aggregation tree inside the candidate clusterheads A minimum aggregation tree-based multihop routing scheme is also investigated which allows the elected cluster heads tochoose the optimally main path to forward data locally and dynamically As a hybridization of the above an evidence-efficientmultihop clustering routing (EEMCR) method is proposed The EEMCR method is simulated validated and compared withsome previous algorithmsThe experimental results show that EEMCR outperforms them in terms of prolonging network lifetimeimproving transmission reliability postponing emergence of death nodes enhancing coverage preservation and degrading energyconsumption

1 Introduction

Decreasing energy consumption improving energy effi-ciency and enhancing transmission reliability are still mainchallenges of wireless sensor networks (WSNs) The relatedtechnique-efficient issues such as clustering routing topol-ogy control and multihop transmission are widely used toimprove energy efficiency for WSNs [1ndash20] On the otherhand it is very important to note that hybridization combi-nation of various approaches may affect total performance

Hierarchical topology control in which nodes aregrouped into clusters and cluster heads (CHs) are elected foreach cluster to form a backbone construction can effectivelyutilize the limited resources of sensor nodes Hierarchicaltopology benefits maximizing the network lifetime andoptimizing the data delivery ratio on each link Clusteringtechnique has been proven energy-efficient in WSNs [3 5ndash7 10 11 13] in which the low-energy adaptive clusteringhierarchy (LEACH) [21] protocol is the most typical oneHowever because of its energy-intensive data transmission

and routing tasks a CH node consumes much more energythan regular sensorsThus an energy-efficientmechanism forCHs rotation or election minimizing energy consumption ofeach node and maximizing the network lifetime while guar-anteeing transmission reliability are still attractive challenges

Multihop routing is one of two main communicationmodes for WSNs In comparison with multipath routingits transmission has generally been considered an efficientenergy-saving approach especially for large-scale sensornetworks [3 7 10 14 16]Themost commonly usedmultihoptopology is the aggregation tree rooted at the sink [22]However the tree topology has an inherent deficiency in thateach sensor has only one path to the sink and the path isnot necessarily optimal which results in that the traffic flowpassing through these sensors may be unbalanced therebymaking some sensors run out of their energy quickly and evenshortening the network lifetime [22] On the other hand theclustering routing algorithms [15 17] integrated the famousDijkstra algorithm and used the Dijkstra algorithm to build

HindawiWireless Communications and Mobile ComputingVolume 2017 Article ID 1914956 14 pageshttpsdoiorg10115520171914956

2 Wireless Communications and Mobile Computing

Sink

0 10 20 30 40 50 60 70 80 90 1000

102030405060708090

100

1

2

3

4

5 67

8 9

10

11

12

13

14

15 16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

4445

46

47

48

49

50

51

52

53

54

55

56

57

5859

60

61

62

63

6465

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

9697

98

99

100

Nodes

CHs

The accumulatedevidenceThe optimum multihopcommunication paths

Figure 1 A demo of the principles employed in the proposedscheme

the shortest path tree with minimum energy consumptionfor each CH to the sink node Namely the CHs couldautomatically form a number of multihop communicationpaths The collected data is continuously transmitted to thesink node via the cluster head-adjacent multihop routing thateffectively shared the overloading of different CHs Howeverthe deficiency of imbalanced distribution of CHs has not beentreated which makes the size of some clusters too large andit is easy to incur the fast energy consumption in some localclusters

In this paper we also address the LEACH-improvedscheme By studying the broadcast nature of data trans-mission in WSNs it is abstracted as a propagation ofresponsibility and availability of nodes The responsibilityand availability integrate sufficient considerations on variousnetwork factors including the residual energy of nodesdistance between nodes distance between the CHs andthe base station and energy loss on the node-joint linksEssentially the responsibility and availability express theaccumulated evidence of nodes to support high-qualitycommunication Namely the responsibility and availabilityrepresent the comprehensive capability of nodes against nodefailures and link losses Based on the above developmentan evidence-efficient CH rotation scheme and algorithmare proposed Furthermore in order to improve the datatransmission reliability between the candidate CHs to thebase station the well-known Kruskal algorithm [23] isemployed to generate the minimum aggregation tree that isa backbone construction algorithm is also developed As ahybridization of the presented subalgorithms an evidence-efficient multihop clustering routing (EEMCR) method isproposed An example of the working mechanism employedin the EEMCR method is demonstrated by several clustersin Figure 1 And more the experimental comparison andanalysis are also examined with the previous algorithms

As far as we know the main contribution of this paperhas at least the following points (1) the eligibility evidenceof a node as a CH is regarded as the sum of responsibilityand availability (eg the accumulated evidence) whichincreases its robustness against random node failures (2)the mechanism of clustering is in terms of the propagation

of responsibility and availability globally which reduces thenegative intercluster communication interference (3) theoptimummultihop communication paths are achieved by thebackbone construction algorithm for the CHs which ensuresthe node-joint link reliability against link failures (4) theassociated algorithms and methods are developed and havevalidated their promising performance

The rest of this paper is organized as follows The relatedworks are introduced in Section 2 In Section 3 we describethe system assumptions and communication models InSection 4 the mechanism and algorithm of CH rotationare examined and proposed We derive the framework ofthe EEMCR scheme and present the overview and thedetailed design of it in Section 5 Promising experimentresults are given in Section 6 and from the effectiveness andefficiency perspective some validations and comparisons areperformed which are followed by the concluding remarksand future works in Section 7

2 Related Works

21 General Clustering Routing Grouping nodes into clustershas been themost popular approach for supporting scalabilityin WSNs [4] Besides the well-known LEACH and LEACH-developed algorithms significant attention has been paid toclustering routingmechanism and algorithms yielding a largeamount of publications [3 5ndash7 10ndash13 17 18 20 24] Theseproposed clustering techniques usually include three phasesCH determination clustering and data transmission Theprevious research works address either one of the issues oroverall three phases fromdifferent perspectives and also theirimprovements and developments with respect to the existingresearches

Usually clustering is typically based on the energy reserveof sensor nodes and nodersquos proximity to the CH [25]Regardless of any of their improvements or developmentthe reduction of energy consumption prolonging the net-work lifetime and improving transmission reliability are thebasic goals in WSNs Energy-efficient clustering techniqueis applied to reduce energy consumption interference andmaintaining connectivity and coverage in WSNs Youniset al [26] proposed a Hybrid Energy-Efficient DistributedClustering (HEED) protocol HEED is a distributed cluster-ing protocol in which a CH election scheme is presentedwith the comprehensive treatment of the residual energyand intracluster communication cost HEED ensures a uni-form distribution of CHs and inter-CH connectivity by anadjustable probability of CH election But the mechanismthat the sensor doubles its probability to become CH duringa repetition phase is not reasonable enough Zhou et al[27] proposed an Energy-Efficient Strong Head clustering(EESH) In EESH nodes are promotedCHs according to theirrespective residual energies their respective degrees and thedistance to and the residual energy of their neighbors Forthat EESH evaluates a cost function for every sensor in thenetwork and iteratively elects the node having the greatestcost as CH This process terminates when all the sensorsin the network are connected to at least one CH Chamamand Pierre [28] proposed a distributed energy-efficient

Wireless Communications and Mobile Computing 3

cluster formation (EECF) protocol EECF elected the CHsfollowing a three-waymessage exchange between each sensorand its neighbors Sensorrsquos eligibility to be elected CH isbased on its residual energy and its degreeThus the messageexchanges complexity of 119874(1) and a worst-case convergencetime complexity of O(N)

Additionally some nonclustering routing methods ormodels [29 30] have also been published They also haveprominent performance on either or all aspects such asdegrading excessive communication energy consumptionprolonging the lifetime and enhancing the transmissionreliability

22 LEACH-Improved Clustering Routing The LEACH pro-tocol one of the first clustering routing protocols proposedfor WSNs is an adaptive distributed algorithm that formsclusters of sensors based on the received signal strength anduses local CHs as routers to the sink node [28] In LEACHeach node has an equivalent opportunity to become CHand through a random rotation of CHs LEACH provides abalance of energy consumption for each node However CHstransmit data directly to the sink node which can be energy-consuming in large-scale WSNs Power-efficient Gather-ing in Sensor Information Systems (PEGASIS) [31] andHierarchical-PEGASIS are two improvements of LEACHUnlike the existing multiple clusters in LEACH PEGASISand Hierarchical-PEGASIS constructed chains of sensornodes so that each sensor node is transmitted and receivedfrom a neighbor and only one node was selected from thatchain to transmit data to the sink node [28] Unfortunatelythe communication between the elected CH and the basestation is one-hop which may waste energy and prove to beunsuitable for large-scaleWSNs [28] Obviously they need tobe improved in-depth

Based on the LEACH algorithm two factors the energyand distance were cast into modifying the threshold func-tion of LEACH protocol a cluster head multihop routingimproved algorithm (CMRAOL) based on LEACH [21] wasproposed in which the multihop communication approachwas adopted and a reliable path was built between the CHsand the sink node the energy consumption of the networkwas effectively balanced but the CH election process wastoo complex Thus the algorithm was only suitable forstatic networks the improved low-energy adaptive clusteringhierarchy-centralized (LEACH-C) algorithm [13] sufficientlyconsidered two factors the energy and number of CHsThe energy consumption of each node improves to be morebalanced but the clustering overhead was too large

Additionally some of the existing clustering routingalgorithms [3 7 10 13 14] did not consider the residualenergy and the deployment location of nodes in responseto this deficiency several researches are carried out in theEBAPC algorithm [18] The EBAPC algorithm defined anew concept of fitness factor and employed the strategy ofcluster center determination inAP algorithm [24]The clustercenter determination strategy [24] was approximated as anelection scheme of the CHs with sufficient consideration ofthe residual energy of nodes In contrast to the previous algo-rithms the CH election was more reasonable and the energy

consumption got more balanced However the EBAPC algo-rithm did not regard the relation between the location ofthe base station and the entire energy consumption of thenetwork which easily leads to high energy consumption ofsome local nodes even premature death of some importantnodes affects the network lifetime and the service quality ofthe network the ELBC and BM-ELBC algorithm [19] basedon EBAPC algorithm were proposed respectively A newconcept of energy level with sufficient consideration of twofactors the residual energy of nodes and distance betweendifferent nodes was introduced the proposed algorithmsmade energy consumption more balanced and they signif-icantly prolonged the network lifetime However the defini-tion of energy level was relatively rough which easily spurredan unreasonable CH election such as taking the remotelyunreasonable nodes as the second level CHs This situationultimately affected the transmission efficiency and trans-mission delay as well as the network lifetime A clusteringrouting algorithm that introduces a new definition of nodecompetitiveness based on it was presented [20] The APBCSalgorithm made the node-joint clustering more uniform andmade the node deployment more reasonable However theAPBCS algorithm had not given an appropriate solutionscheme to satisfy the requirement of multihop transmissionscheme within the node-joint CHs as a result increasing theenergy consumption on some CHs When the scale of WSNsgets large it leads to the occurrence of a premature death forthe CHs such that it was unable to guarantee the coveragepreservation and network connectivity

The differences covered between the proposed EEMCRscheme and the previous researches include the following thenature of data transmission in WSNs is abstracted as propa-gation of responsibility and availability and the responsibilityand availability are taken as the accumulated evidence ofnodes for guaranteeing transmission reliability the responsi-bility and availability integrate comprehensive considerationson several network factors an evidence-efficient cluster headrotation mechanism and algorithm are presented backboneconstruction algorithm is developed to generate a minimumaggregation tree essentially the branches inside the gen-erated minimum aggregation tree construct the multihopcommunication paths The experimental results show thatthere exists a remarkable improvement on prolonging thenetwork lifetime postponing the death of nodes savingenergy and conducting coverage preservation

3 Assumptions and Models

31 System Assumptions Assuming that 119899 nodes are ran-domly deployed in a target area in which each node hasthe same configuration and data processing capability thelocation of each node is known-available Additionally thispaper makes the following assumptions(1) All the sensor nodes in the system have the sameorganization and initial energy 1198640(2) All sensor nodes adaptively adjust the transmissionpower according to their need


Amplifiercircuit

Transmittingcircuit

Receivingcircuit

d

l bit data l bit data

EＦ＝ times l EＦ＝ times lＧＪ times l times dn

E２８(l d)ET８(l d)

Figure 2 Wireless transceiver circuit energy consumption model

32 Communication Model In this paper a well-knowncommon first-order wireless energy consumptionmodel [32]is employed which is shown in Figure 2

With respect to the abovemodel the energy consumptionof transmitting l bit data is composed of transmission circuitpower and power amplification losses Different power con-sumption models are employed with different distances andthe energy consumption of transmitting l bit data by nodescan be calculated by the following formula

119864TX (119897 119889) = 119864TX elec (119897) + 119864TX amp (119897 119889)=

119897119864elec + 119897120576fs1198892 119889 lt 1198890119897119864elec + 119897120576amp1198894 119889 ge 1198890

(1)

where 119864elec represents the energy consumption by a circuitprocessing a single bit data 120576fs denotes the coefficient ofpower amplifier in the free spacemodel 120576amp is the coefficientof multidiameter attenuation model of power amplifier and1198890 is the critical distance between free space propagationmodel and multipath attenuation model and is calculated asthe following formula

1198890 = radic 120576fs120576amp (2)

The energy consumption by the node receiving 119897 bit datais formalized as the following formula

119864Rx (119897) = 119864Rxminuselec (119897) = 119897119864elec (3)

4 Evidence-Efficient Cluster HeadRotation Mechanism

In this section we discuss the issue of CH rotation Firstwe think that the essence of CH election in clusteringtopology is equivalent to determining the cluster centroidsof cluster algorithms in data mining The Affinity Propa-gation (AP) [24] cluster algorithm is very prominent at itsperformance on carrying out large amount of data Thuswe employ the propagation mechanism in AP algorithmand study the scheme of CH rotation associated with it

Under sufficient consideration of several network factorsincluding the residual energy of nodes distance betweennodes energy consumption associated transmission distanceand amount of data and abstracting the capability of guaran-teeing transmission reliability in WSNs as the responsibilityand availability propagation between node-joint links themaximum sum of responsibility and availability is taken asthe accumulated evidence of CH rotation Namely if theaccumulated evidence of a node is relatively large then ithas a relatively high possibility of becoming a CH which isin charge of maintaining the node-joint links and ensuringtransmission reliability

Additionally since the energy efficiency is one of the crit-ical factors that influence the network lifetime in particularseveral new definitions with respect to the measurement orestimation of energy consumption are presented as follows

Definition 1 Node-Energy-Level (NEL) is used to measurethe energy consumption of nodes in the network NEL iscalculated as Formula (4)

NEL = (119864remain

119864init )1119873live (4)

where 119873live denotes the number of surviving nodes in thecurrent status of the network 119864init represents the initialenergy of a node 119864remain denotes the residual energy ofa node The meaning of NEL shows that the higher theresidual energy of a node the greater the value of the nodeenergy level (eg NEL) which indicates that the stronger thecurrent node activity the greater the coverage preservationas well as the better the quality of service (QoS) In contrastwhen the residual energy of nodes is lower the node energylevel is smaller which indicates that the smaller coveragepreservation and the worse QoS

Definition 2 The Energy-Cost EC(119894 119895) is defined as thefollowing formula

EC (119894 119895) = ET (119894 119895)119864remain119894

(5)

where 119864remain119894 denotes the current remainder energy of CH i

and ET(119894 119895) represents the real amount of energy consumedbyCH 119894 to transmit a unit data to CH j EC(119894 119895) only expressesa logical transmission overhead rather than a real metric ofenergy consumption Itsmeaning not only shows consideringthe energy consumption of the real data transmission butalso takes into account the residual energy level of nodesthemselves

Definition 3 119904(119894 119895) indicates the level of node j appropriatefor the cluster head to node i which is called the orientation-tended matrix and defined as the following formula

119904 (119894 119895) = minus(120572 lowast (119889 (119895 119861) lowast 119864remain

119895 )NEL + (1 minus 120572) lowast 119889 (119894 119895)) 119894 = 119895 120572 isin [0 1]119875 (119895) 119894 = 119895 (6)


whereB represents the base station119889(119894 119895) is the distance fromnode i to node j and with the Euclidean distance to express119889(119895 119861) is the distance between node j and base station119864remain

119895

is the current residual energy of node j NEL is the Node-Energy-Level (eg NEL) of the current node and is calculatedusing Formula (4) 120572 is a weight used to adjust the residualenergy and distance in the CH election scheme Usually anode with a larger value of 120572 and more remainder energyand closer to the base station is easier to become a CH 119875 isthe orientation-tended parameter that is the diagonal valuein the orientation-tended matrix defined as the followingformula

119875 (119894) = minusNEL times ( 119864init119894119864remain119894

) (7)

where 119864init119894 is the initial energy of node i It can be seen from

Formula (7) because the orientation-tended matrix is alwaysnegative Thus the more the residual energy of the node thegreater the value of the orientation-tended parameter 119875 aswell as the higher the probability that the node will becomethe CH

Based on the above this paper integrates inspirationson the broadcast nature of data transmission in WSNs andthe idea in AP algorithm Furthermore we give improve-ment recognition of the responsibility and availabilityrespectively

Definition 4 The eligibility evidence of a node as a clusterhead is quantified using the responsibility and availabilityThey are defined as follows

119903 (119894 119896) = 119904 (119894 119896) minusmax 119886 (119894 1198961015840) + 119904 (119894 1198961015840)(1198961015840 isin 1 2 119899 1198961015840 = 119896) (8)

119886 (119894 119896) = min0 119903 (119896 119896) +sum1198941015840

max 0 119903 (1198941015840 119896)(1198941015840 isin 1 2 119899 1198941015840 = 119894 1198941015840 = 119896)

(9)

where 119903(119894 119896) is the responsibility between node i and nodek If k is taken as the potential CH then responsibility istransmitted from node i to node k and is equivalent to theaccumulated evidence for node k to be the CH of node i119886(119894 119896) as the availability between node i and k transmits fromnode k to node i It implies the accumulated evidence fornode i to select node k as its CH Thus it measures whetherk can finally become the real CH after each cycle of self-adaptation

Formulas (8) and (9) iterate according to Formulas (10)and (11) respectively

119903new (119894 119896) = (1 minus 120582) 119903 (119894 119896) + 120582119903old (119894 119896) (10)

119886new (119894 119896) = (1 minus 120582) 119886 (119894 119896) + 120582119886old (119894 119896) (11)

where 120582 isin [0 1] is the learning rate in updating 119903(119894 119896) and119886(119894 119896) At the beginning of the iteration their initial valuesare set to ldquo0rdquo

Usually link losses and node failures are the primaryreasons to influence transmission reliability in WSNs unfor-tunately they are negatively affected bymanynetwork factorsIn this paper we take the availability and responsibility as thecomprehensive evidence against node failures and link losseswhich aims to improve the total performance includingtransmission reliability and other properties To this pointthe availability and responsibility provide a logical metric fornodes to guarantee transmission reliability

Furthermore in order to ensure the rationality of CHrotation and correctness of CH election mechanism we givethe following theorem and proof

Theorem 5 The greater the sum of responsibility and avail-ability the greater the possibility of a node to become a clusterhead

Proof The essence of a ldquoclusterrdquo in the clustering routingis similar to the ldquocluster analyzerdquo in the associated clusteralgorithms of datamining thus the CH election likes the pro-cedure of determining the cluster centroids in the algorithmsof dataminingMoreover the well-knownAP algorithm is anessential cluster algorithm and its cluster procedure is carriedout by the propagation of responsibility and availability inwhich the first critical business is to adaptively determinethe different cluster centroids according to the changeableorientation parameter The cluster procedure is iterativelyongoing until all of the data items are clustered This aboveidea is equivalent to the mechanism of the CH rotation inWSNs and equivalently applied in selecting the CHs that isthere exists a set of dynamic candidate CHs composed of alarge amount of nodes which correspond to the cluster cen-troids and are traversed by the propagation of responsibilityand availability In each cycle of the clustering the node-joint links build the interpath within a single cluster and thecluster head-joint links set up the intrapath between differentclusters During the propagation processing the availabilityand responsibility denote the capability of resisting nodesfailures and link losses as well as their propagation directionwhich guides the data transmission paths Obviously thegreater the sumof responsibility and availability is the greaterthe probability that a node becomes a CH would be Thetheorem is proved

Consequently in a real clustering phase each cycleincludes two phases cluster head election and clusteringThe two phases are periodically ongoing until any noderuns out of its energy The responsibility and availabilitytransmit themselves within nodes-joint links in the clusteringprocedureThe sumof ldquo119903(119894 119896)+119886(119894 119896)rdquo is adaptively changingThe larger the sumof 119903(119894 119896)+119886(119894 119896) the greater the probabilitythat the node k becomes aCH otherwise the node k is unableto become aCHNamely as for node 119894 it always selects node 119896that maximizes the sum of 119903(119894 119896) + 119886(119894 119896) Furthermore if 119894 =119896 node 119894 is the CH otherwise 119894 = 119896 node 119896 is the CH of nodei In combination with Definitions 1 2 and 3 and the above-mentioned analysis it indicates that the nodes with moreresidual energy and little amount of data to transmit have ahigh probability to be CHs Apparently the CHs election isan adaptive and dynamic process


InputMax iterations IterMaxOutput Candidate cluster head nodes(1) Calculate NEL according to Formula (4) and send the NEL value to sink node(2) Sink node calculate the 119878 and 119875 according to Formula (6) amp (7) respectively(3) Sink node broadcast the 119878 and 119875(4) Set119872119886119909 = 0(5) while (Current iterations 119868119905119890119903 lt 119868119905119890119903119872119886119909)(6) Calculate 119903(119894 119895) and 119886(119894 119895) according to Formula (8) amp (9) amp (10) amp (11)(7) Obtain node k with max(119886(119894 119896) + 119903(119894 119896)) for node i(8) if 119894 = 119896(9) Node i is the cluster head(10) else if 119894 = 119896(11) Node k is the cluster head of node i(12) end(13) Current iterations 119868119905119890119903 = 119868119905119890119903 + 1(14) end

Algorithm 1 Clustering

Lemma 6 The propagation direction of responsibility andavailability as themultihop transmission path effectively avoidthe node failures

Proof Use reduction to absurdity to prove it From theperspective of the definition of responsibility and availabilityas long as a node fails in the network its availability auto-matically gets ldquo0rdquo and the associated responsibility of it alsobecomes ldquo0rdquo that is the responsibility or availability does notcontinuously spread along such a path including these nodesThus the lemma is proved

Lemma 7 The termination condition for the propagation ofresponsibility and availability is the energy exhaustion of anynode in the network

Proof As it can be seen from the aforementioned Definitions1 2 3 and 4 once the energy of a node in the network getsldquo0rdquo then its corresponding orientation-tended parameterbecomes ldquo0rdquo As long as the orientation-tended parameter isldquo0rdquo the node is unlikely to be elected as the CH in the nextcycle of clustering Similarly we can see that once the energyof all nodes in the network gets ldquo0rdquo the transmission of theresponsibility and availability is automatically terminated

Evidently the aforementioned description is a globalelection scheme for CH rotation Namely all of the runningnodes also have the same opportunity to become candidateCHs regardlesswhether they used to beCHs or notUndoubt-edly this scheme is helpful to achieve balance in energyconsumption of each node and prevent the premature deathof any node thereby guaranteeing coverage preservation andprolonging the network lifetime

5 EEMCR Routing Scheme

In this section we talk about the proposed EEMCR routingscheme in detail Exactly EEMCR scheme inherits the basicframework of LEACH algorithm which hybrids the basic

processing of clustering and data transmission within eachcycle In fact EEMCR scheme integrates several subalgo-rithms on CH rotation and backbone construction suchhybridization subalgorithms are ongoing in an iterationmode and they finally complete multihop clustering routingThe details of each subalgorithm are discussed in the follow-ing parts respectively

51 Clustering The EEMCR conducts the clustering in themechanism of iteration cycle by cycle which is similar tothat in LEACH algorithm However the LEACH protocolrandomly replaces the CHs in each cycle of iteration incontrast the EEMCR method adaptively updates the CHsaccording to the accumulated evidence of each node In theinitial stage of each cycle for the proposed EEMCR the basestation calculates the orientation-tended matrix of the nodesin terms of the Formula (6) and broadcasts it to all nodesThenodes that successfully receive the orientation-tendedmatrixcalculate their own responsibility and availability accordingto Formulas (8) and (9) At the same time for any node i thealgorithm takes the node k thatmaximizes the sum of 119903(119894 119896)+119886(119894 119896) as the CH in the current cycle of iteration Along withthe ongoing iteration each node updates its responsibilityand availability according to Formulas (10) and (11) until thedemands of iterations or the convergence of the algorithm isachieved

Summarizing the above steps describes the clusteringalgorithm as shown in Algorithm 1

Theorem 8 The time complexity of clustering algorithm is119874(119899)Proof The termination condition of Algorithm 1 is either thenumber of iterations exceeds the set threshold or the CHswithout any ongoing changes Obviously there is no negativeimpact on the complexity of the algorithm Therefore thetime complexity of Algorithm 1 is also 119874(119899)


(1) Qlarr 0 lowast Initialization Q is the set of theminimum aggregation tree(2) for each path 119890(119894 119895) isin 119866 lowastG represents the node set of the candidate cluster heads(3) Sort the path 119890(119894 119895) into an increasing order by weight EC(119894 119895)(4) end(5) for each 119890(119894 119895) isin 119866 taken in an ascending order by weight EC(119894 119895)(6) if 119890(119894 119895) notin 119876ampampNo loop then(7) Qlarr 119876 cup 119890(119894 119895)(8) end(9) end(10) return Q

Algorithm 2 Backbone construction

52 Backbone Construction Backbone construction essen-tially consists of generating minimum aggregation tree andminimum aggregation tree-based multihop communicationpaths according to graph theory On the other hand the mul-tihop transmission path selection problem is very complexThe following theorem is given firstly

Theorem 9 Multihop routing with respect to the cluster headsin WSNs is an NP-complete problem

Proof At the stage of building the minimum aggregationtree the energy consumption of each CH depends on itsEC (eg Definition 2) with its neighbor CHs It can beseen from Definition 2 that the EC of a CH is determinedby its residual energy and its distance to the neighbor-adjacent CHs Therefore in order to minimize the energyconsumption and improve the transmission reliability weneed to find a spanning tree so that the number of its neighborCHs for eachCH is the smallest and the interdistance betweenthe different CHs is the shortest Namely it is necessary tofind a minimum spanning tree Garey and Johnson [33] haveproved that the minimum spanning tree problem is an NP-complete problem

Additionally the well-known Kruskal algorithm in thefield of graph theory is suitable for solving the NP-completeproblems [23] In this paper we also employ it to conductthe issue of multihop routing selection within CHs to tryto achieve an approximate solution for it In fact the goalis to construct the connected graph with Kruskal algorithmin which the selected child nodes from the candidate CHstake the value of the EC as the vertexes (eg Definition 2)between a pairwise of the selected nodes and the weight onthe corresponding edge Finally the essence of the multihopcommunication path construction is to generate an aggrega-tion tree from the connected graph of which the edge with arelative minimum weight is the criteria for selection namelyselecting the edges from the graph to build a tree accordingto the ascending order of the weight Once a single edge witha relatively small weight is selected and the currently selectededge with the previously selected edges does not form a loopthen it is preserved in the generated tree otherwise the latestselected edge is removed As a result it finally obtains a treewith 119899clusterminus1 items of edgesThus the theorem is proved

Furthermore the construction of multihop communi-cation paths within the candidate CHs is as follows firstusing the proposed Algorithm 1 to determine the candidateCHs second the selected CH nodes deliver a status messageto the base station The status message is expressed as aframe of 119899119900119889119890 119868119863 119899119900119889119890 119903119890119904119894119889119906119886119897 119890119899119890119903g119910 119897119900119888119886119905119894119900119899 thirdthe base station calculates the EC between the CHs basedon the received status messages and broadcasts the obtainedEC information to all of the CHs finally the CHs that havereceived the broadcasting message from the base stationcooperatively construct an aggregation tree whose root nodeis the base station Consequently the data can be transmittedto the base station with a multihop transmission approachalong the paths inside the generated aggregation tree theforwarded data get to the base station along the branchesfrom the leaves to the root inside the generated aggregationtree Essentially the built aggregation tree is the backboneconstruction

In summary backbone construction algorithm is pre-sented as shown in Algorithm 2

Theorem 10 The generated tree conducted by the Kruskal-based method is a minimum aggregation tree

Since the typical Kruskal algorithm targets to find anundistorted minimum aggregation tree thus we use thereduction to absurdity to prove theTheorem 10

Proof Assuming that there exists a really minimum aggre-gation tree 1198791 in the network in contrast 1198792 is anotheraggregation tree obtained by the Kruskal algorithm and 1198791 =1198792 thus there is at least one path e that falls inside 1198792 ratherthan 1198791 If the path e is removed from 1198792 1198792 becomes twoparts of nonconnected subtrees which are denoted by 119879119886 and119879119887 respectively On the other hand there must be a path 119891existing in 1198792 of which one of the vertex of 119891 is located in119879119886 and the other one is located in 119879119887 satisfying the conditionEC(119891) lt EC(119890) Furthermore it is impossible to select path edue to the EC(119891) lt EC(119890) in contrast the path119891 is preferablyselected according to the Kruskal-associated aggregation treegeneration method Under the worst situation if the path eis forced to join the connected subsets the loop inevitablyemerges This result is in contradiction with the basic idea of


(1) Initializing network parameters(2) Sink node collects all the related parameters of the nodes in the network(3) if sink node receives all the relevant parameters of the nodes in the network(4) call Algorithm 1(5) call Algoithm 2(6) end(7) Sink node broadcasts cluster head information and minimum aggregation tree information(8) for each node i received the cluster head information and minimum aggregation tree information(9) if (node i is the CH)(10) Find the next hop (eg cluster head) from the clues in the minimum aggregation tree(11) else(12) Find its cluster head from the candidate cluster heads(13) end(14) end(15) for each member of a cluster in the network(16) Performing data transmission by single-hop mode(17) end(18) for each CH in the network(19) Performing data forwarding(20) end

Algorithm 3 EEMCR method

the Kruskal algorithm Namely only when 1198791 = 1198792 holdsTheorem 10 is correct Proof is completed

Based on the above analysis and proof under the consid-eration of building a new tree denoted as 1198793 1198793 satisfies thefollowing formula

1198793 = 1198792 minus 119890 + 119891 (12)which implies that using path 119891 connects 119879119886 and 119879119887 Incontrast 1198793 is the finally generated aggregation tree using theKruskal-based method It is easy to find that the EC of 1198793 isless than that of 1198792 which implies retaining the edge with asmall weight of EC and removing the edge with a large weightof EC Similarly as for m different paths which exist in 1198791and 1198792 the minimum aggregation tree 1198791 can be successfullyachieved by m times of the transformations like the abovesteps Namely the agglomeration trees in the network canbe transformed into the minimized aggregation tree by theKruskal-like algorithm that is the minimum aggregationtree 1198791 must be available

53 Evidence-Efficient Multihop Clustering Routing SchemeHybridizing the aforementioned Algorithm 1 Algorithm 2and some necessary steps are described as an evidence-efficient multihop clustering routing (EEMCR) method asshown in Algorithm 3 In Algorithm 3 to the intraclustercommunication the CH and its members communicatedirectly with single-hop transmission mode whereas themultihop communication approach is examinedwithin inter-clusters

Theorem 11 The time complexity of EEMCR scheme is119874(119899 log 119899)Proof Assuming the connected graph 119866(119881 119864) with 119899119905 ver-tices and 119899119890 edges using results from the process of building

the minimum aggregation tree according to the Kruskalalgorithm and the weight of each edge is the correspondingEC (eg Definition 2) Firstly since the Kruskal algorithmbuilds a subgraph with only 119899119905 vertices without any edge atthis time this subgraph can be equivalently regarded as aforest with 119899119905 items of trees and the vertices in the subgraphare taken as the root node of each tree respectively Thenthe Kruskal algorithm selects an edge from the candidate setof edges in the network with the relative smallest weightand if the two vertices of the selected edges belong to twoitems of different trees the two pieces of trees are combinedas a new tree by connecting the two vertices that is theselected edge is added to the subgraph In contrast if thetwo vertices of the selected edge have fallen into the sametree the selected edge is not desirable instead the next edgewith the relative smallest weight is tried continuously Thissimilar process continues until there remains only one treein the forest that is the built subgraph contains 119899119905 minus 1 itemsof edges Consequently EEMCR scheme just scans the 119899119890items of edges at most once and selecting the edge with theminimum EC requires only the time of 119874(119899 log 119899) Thus thetime complexity of Kruskal-basedminimum aggregation treegeneration method is 119874(119899 log 119899)

In summary since the time complexity of clustering is119874(119899) and the time complexity of backbone constructionis 119874(119899 log 119899) the time complexity of EEMCR method is119874(119899 log 119899)6 Experimental Analysis and Comparison

In this section several experiments are arranged to validatethe effectiveness and efficiency of the proposed methodfrom different aspects The parameters configured in thesimulation experiments are shown in Table 1 All of theexperiments are realized using Matlab R2010b


0 50 100 1500

102030405060708090

100

(a) The clustering results using EEMCR0 50 100 150

0102030405060708090

100

(b) The clustering results using LEACH

Figure 3 The clustering results

Table 1 Parameters in the experiments

Parameter type Parameter valueNode distribution area 100m times 100mSink location (150 50)Node numbers 100sim850Initial energy of sensor node 1198640 03 JCircuit processing data consumption energy 119864elec 50 nJbitPacket length 2000 bitsControl packet length 32 bits120576fs 100 pJbitm2120576amp 00013 pJbitm4

Data aggregation energy consumption 5 nJbitsignalData aggregation rate 06120582 0IterMax 1000120574 09

61 Effectiveness Figures 3(a) and 3(b) are the experimentalresults of clusters using EEMCR and LEACH algorithmsrespectively The horizontal and vertical coordinates of thegraph represent the location information of the nodes ldquo0rdquorepresents an ordinary node and the sinkrsquos location is (15050) The noncluster head nodes in Figure 3(a) are connectedto the CH with a red dashed line and the CH node isconnected to the sink with a dark blue dashed line thenoncluster head node in Figure 3(b) is connected to theCH with a solid line and the CH is connected to the sinkwith a dashed line Obviously due to EEMCR scheme takinginto account several network factors the distribution of thedetermined CHs is reasonable and the scale of clusters isrelatively appropriate in contrast the CH distribution usingLEACH algorithm is random and the cluster size is too large

Furthermore to validate the influence of the proposedclustering and backbone construction algorithm on theperformance we fundamentally implement the LEACH[21] LEACH-C [13] and CMRAOL [14] algorithms respec-tively Additionally through casting the proposed CH rota-tion mechanism and minimum aggregation tree generationinstead of the original ones in the LEACH algorithm the

Table 2 Comparisons of the five algorithms

Name FirstDeath LifeTime MaxEC MinECLEACH 24 301 00167 00011LEACH-CS 65 377 00074 00015LEACH-C 43 345 00141 00007LEACH-MT 36 326 00117 00009CMRAOL 33 313 00150 00021

hybrid results are expressed as LEACH-CS and LEACH-MT respectively Namely embedding the proposed clus-tering algorithm replaces the clustering phase in LEACHand generates the LEACH-CS method the backbone con-struction algorithm builds the communication path insteadof the original one in LEACH the generated combinationmethod is denoted as LEACH-MT So the LEACH LEACH-C and LEACH-CS are associated with the LEACH algo-rithm and with or without considering the proposed clusterhead rotation mechanism whereas the LEACH-MT andCMRAOL algorithms consider the multihop transmissionapproach Finally the effectiveness is validated on severalindices including the network lifetime (LifeTime) emergenceof first death node (FirstDeath) maximum average energyconsumption (MaxEC) and minimum average energy con-sumption (MinEC) of nodes The experimental results areshown in Table 2

Regarding the first emergence of death nodes in Table 2the first emergence time is during the 65th cycle of clus-tering in LEACH-CS that is far later than that of theother compared algorithms LEACH-C is the second bestwherein the LEACH algorithm first causes the death nodeto emerge at its 24th cycle and is the worst This is becauseLEACH-CS employs the proposed evidence-efficient clusterhead rotation mechanism to elect the CHs and achievesa significantly delayed energy consumption of each nodeand effectively prolongs the network lifetime Although theLEACH-C algorithm takes the energy factor into accountin the CH election stage it consumes more energy in theclustering stage which leads to the premature death of thecluster nodes compared to the LEACH-CS However the firstemergence of death nodes in LEACH-C compared with that


in LEACH algorithm is improved the cause reason is becauseLEACH-C utilizes a cluster head election strategy againstthat of randomly selecting the CHs in LEACH algorithmThe impropermechanism in LEACH results in insignificantlylarge amounts of energy consumption

The network lifetime mainly reflects the capability ofdifferent algorithms to allocate the energy and schedulethe task of data transmission As shown in Table 2 thenetwork lifetime of LEACH-CS algorithm is the longest andthe network lifetime of LEACH-C algorithm is the secondlongest The network lifetime of these two algorithms greatlyimproved in comparison with that of LEACH algorithmThe main reason is that a relatively proper cluster headelection is taken in them which results in slowing energyconsumption and effectively prolonging their lifetime Onthe other hand the lifetime of LEACH-MT and CMRAOLalgorithms is relatively close and also longer than that of theLEACHalgorithmThe primary reasons are that the LEACH-MT andCMRAOL employ amultihop communicationmodeand leverage link redundancy to improve data transmissionefficiency and energy consumption However in comparisonwith the LEACH-CS and LEACH-C the LEACH-MT andCMRAOL only lead to a little achieved improvement forthe network lifetime since the CH election is not entirelyreasonable in the clustering stage in which the clusteringprocess still consumes a large amount of energy against theLEACH-CS algorithm

The maximum average energy consumption and min-imum average energy consumption of the nodes is ana-lyzed The difference between the maximum average energyconsumption and minimum average energy consumptionfor the nodes is regarded as the index a larger differenceindicates the more energy consumption is centralized on afew nodes with a large amount of workload in contrast asmaller difference indicates the energy consumption and itsdistribution is more reasonable and uniform which benefitsthe coverage preservation From Table 2 the difference ofaverage energy consumption using LEACH-CS is the smallestone of 00059 the difference of average energy consumptionusing LEACHalgorithm is the largest one of 00156 Althoughthe LEACH-MT and CMRAOL algorithms employ multi-hop communication approach their difference of averageenergy consumption also reaches more than 001 The aboveresults indicate that LEACH-CS achieves a promising balancebetween energy consumption of nodes thus prolonging thenetwork lifetime and enhancing coverage preservation

In summary the hybridization LEACH-CS methodshows the best performance in terms of the indices LifeTimeFirstDeath MaxEC and MinEC The experimental resultsindicate that the proposed cluster head rotation mechanismeffectively limits the transmission range and set of neighborsof nodes

62 Efficiency In this part experimental comparisons andanalyses are performed between the EEMCR method andother methods such as the LEACH-developed algorithmincluding BM-ELBC [19] ELBC [19] and EBAPC [18] aswell as LEACH [21] itself on the indices including the totalenergy consumption of networks survival time of nodes

0 50 100 150 200 250 300 350 4000

102030405060708090

100

Cycles

Num

ber o

f dea

th n

odes

LEACHEBAPCELBC

BM-ELBCEEMCR

Figure 4 The death nodes versus the cycle of clustering

average energy consumption of nodes and influence of thesink location which are to show its efficiency

Usually the emergence of death nodes is later andnumberof death nodes is less which indicates higher coverage preser-vation less communication holes better energy efficiencyand higher QoS Experiment 1 clearly shows the effect thatthe proposed EEMCR method has on the emergence ofdeath nodes The results of the experiment are presented inFigure 4 As shown in Figure 4 the emergence of death nodeswith EEMCR is the latest one and BM-ELBC algorithm isthe second latest In particular the EEMCR method is farlater than that of the other compared algorithms When thecycle gets 400 almost half of the nodes are also survivingfor the proposed EEMCR method it is very prominentamong the compared algorithms The main reason for thissituation is that EEMCRmethod examines an effective clusterhead rotation mechanism which effectively avoids the rapidenergy consumption for the CHs thereby slowing the deathof nodes and enhancing the coverage preservation On theother hand since the EEMCR algorithm is based on themultihop transmission mode associated with the minimumaggregation tree-based generation methods which buildsnode-joint paths with the smallest EC (eg Definition 2)between the CHs that is the optimum path consequentlythe energy consumption in the process of forwarding data tothe base station significantly degrades Namely it improvesthe energy efficiency and reduces the monitoring blind areaAlthough the BM-ELBC algorithm also employs the multi-hop communicationmode it only considers the transmissioncapability rather than the energy losses on the multihopcommunication path according to the residual energy of theCHs This case incurs fast energy consumption for the CHswith a higher opportunity of carrying out the data transmis-sion tasks and leading to premature death of these CHs Incontrast LEACH and ELBC algorithms firstly generate deathnodes and the EBAPC algorithm is the second This is firstlybecause LEACH ELBC and EBAPC algorithms are onlybased on single-hop communication mode Additionally theCHs election is relatively unreasonable in LEACH protocol


0 50 100 150 200 250 3000

5

10

15

20

25

30

Cycles

LEACHEBAPCELBC

BM-ELBCEEMCR

The t

otal

resid

ual e

nerg

y (J

)

Figure 5 The residual energy versus the cycle of clustering

Additionally EBAPC algorithm is not the algorithmwhose emergence of death nodes is the earliest one but itsuptrend gradient is the largest one that is its ratio of deathnodes is the highest The primary reasons which caused thisphenomenon are that the CHs have to undertake more dataforwarding tasks after the cluster formation which spur a fastenergy consumption on some nodes thus it is not even goodas that of LEACH algorithm

Experiment 2 demonstrates how the algorithms performin relation to the speed of total energy consumption in thenetwork Generally the speed of total energy consumptionis one of the key indices to evaluate the comprehensiveperformance of networks The more and faster the totalenergy consumption is the shorter the survival time of nodesis The experimental results are shown in Figure 5 It can beseen from Figure 5 that the descending trend of the totalenergy consumption of EEMCR method is the slowest theBM-ELBC algorithm is the second slowestThey are superiorto that of LEACH and EBAPC algorithm This is becauseEEMCR and BM-ELBC algorithms use the multihop trans-mission approach whereby leveraging the link redundancy toimprove the transmission reliability and balance the energyconsumption of CHs as well as slowing down the speedof energy consumption for the whole network Moreoverin comparison with the BM-ELBC algorithm the EEMCRalgorithm has outperformed it on saving energy This ismainly because the optimum CH election and the selectedclusters head always choose the best multihop routing forforwarding their aggregation data to the base station therebyachieving an effective reduction of the insignificant energyconsumption In contrast LEACH and EBAPC likely choosethe nodes with less residual energy as CHs in the procedureof clusteringwhich causes fast energy consumption and leadssome nodes to premature death

Experiment 3 shows how the average energy consump-tion of nodes is affected by the algorithms The experimentalresults are shown in Figure 6 Generally the average energyconsumption of nodes fully reflects equilibrium level ofenergy consumption and also shows the optimization levelof the selected multihop routing for data forwarding from

0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

7

8

ID of nodes

LEACHEBAPCELBC

BM-ELBCEEMCR

The a

vera

ge en

ergy

cons

umpt

ion

(J)

times10minus3

Figure 6 The average energy consumption of different nodes

another aspect Under the same premise the lower theaverage energy consumption of nodes is the longer the nodesurvives the stronger the coverage preservation is and themore reasonable the forwarding routing selection is Thecloser the average energy consumption of each node themore similar the energy consumption of different nodessuggesting the self-organization and multihop transmissionpath of each node in WSNs are more reasonable and theenergy consumption is more balanced which may lead toa closer survival time of each node and more integrity ofthe network connectivity As shown in Figure 6 the averageenergy consumption of nodes with the EEMCR algorithm isthe most stable and is apparently lower than that of othercompared algorithms The BM-ELBC algorithm shows thesecond best These experimental results indicate that thetransmission paths using EEMCR algorithm are perfect theenergy consumption among different nodes achieves balanc-ing the survival time of the nodes is close to each other andthe capability to guarantee transmission reliability in WSNsis relatively strong This is due to the EEMCR algorithmemploying the evidence-efficient cluster head rotation mech-anism to examine the clustering procedure which results in arational distribution of clusters and a proper size of clustersas well as a balanced energy consumption among differentnodes Besides this the selected minimum aggregation tree-based multihop routing using EEMCR algorithm leveragesthe link redundancy to improve transmission reliability andenergy efficiency Although the EBAPC algorithm also showsrelatively stable average energy consumption among differentnodes its fluctuation is stronger than that of EEMCR algo-rithm and weaker than that of the remaining compared algo-rithmsThe spurred reason is that the EBAPC algorithm alsoreasonably assigns the data forwarding tasks and dispersesthe energy consumption in WSNs to a certain extent Incontrast LEACH and ELBC algorithms show a strong fluctu-ation of the average energy consumption on some nodes Forexample with the combined analysis shown in Figure 3(b)


100 110 120 130 140 150 160 170 180 190 2000

50100150200250300350400450

Location of the sink node

LEACHEBAPCELBC

BM-ELBCEEMCR

deat

h no

deCy

cles u

ntil

the fi

rst

(a) The emergence of the first death node versus the changing location ofthe sink

100 110 120 130 140 150 160 170 180 190 200100

200

300

400

500

600

700

800


LEACHEBAPCELBC

BM-ELBCEEMCR

Life

time

(b) The survival time versus the changing location of the sink node

Figure 7The influence of base station location on the survival time

since the size of clusters is unreasonable using LEACHalgorithm some CHs have to undertake very large amountsof data transmission which results in the distinguishedfluctuation in energy consumption On the other handthe average energy consumption between different nodes isvery large which indicates that the energy consumption iscentralized on a few nodes and the energy consumption isunbalanced the distribution of nodes shown in Figure 3(b)also verifies this point Like that of LEACH algorithm thesimilar fluctuation also happens in the ELBC algorithm theprimary reasons are due to the unreasonable CH electionand lead to a tremendous variation of energy consumptionfor some local nodes In summary Figure 6 shows thatthe generated multihop communication path using EEMCRalgorithm is reliable data forwarding efficiency is relativelyhigh thus not only slowing the overall energy consumptionbut also globally balancing energy consumption of differentnodes

Experiment 4 shows how changing the base stationlocation affects the algorithm performance Changing thelocation of the base station can better evaluate the robustnessof the algorithm to the variation environment The experi-mental results are shown in Figure 7 The base station moves

0 100 200 300 400 500 600 700 800 900100

200

300

400

500

600

700

800

Node number

Life

time

LEACHEBAPCELBC

BM-ELBCEEMCR

Figure 8 The influence of the scale on algorithms

from (100 50) to (200 50) namely taking shifting horizontalcoordinate as an example The distance between the basestation and the target area gets farther and farther theemergence of the first death node with the five comparedalgorithms is getting more and more early and the lifetimegets shorter and shorter With respect to the descendingtrend the EEMCR algorithm is the slowest which indicatesthat the emergence of the first death node is effectivelypostponed At the same time this phenomenon indicates thatEEMCR algorithm has a better robustness to adapt to thechanging configuration this is mainly because it employs theefficient cluster head rotation mechanism while choosing abestmultihop routingAndmore since considering the resid-ual energy of nodes it degrades the possibility of theCHswithlower residual energy as the intermediate forwarding nodesand significantly avoids their rapid energy consumption andeven causing the premature death of the nodes undertakingforwarding tasks Additionally when the base station isfarther the BM-ELBC algorithm also delays the occurrenceof the first death node but it is still earlier than that ofEEMCR algorithm this is mainly because the BM-ELBCalgorithm failed to choose the optimum communicationpathwhich easily leads to increasing the energy consumptionand shortening the network lifetime

Experiment 5 demonstrates the algorithm capability toadapt to large-scale WSNs and the nodes changes betweenthe ranges of 50ndash850 The experimental results are shownin Figure 8 in which the lifetime is taken as the primaryevaluation index

Usually as the scale of the network increases that isthe number of nodes in the network becomes larger theworkload also increases accordingly and thus theCHelectiongets more critical Figure 8 shows the network lifetimechanges with the increasing number of nodes from 50 to 850Especially after the number of nodes is greater than 200 thelifetime of the proposed EEMCR scheme is more prominentthan that of the others and the survival time of the networkincreases greatly This situation indicates that the proposedEEMCR algorithm can effectively prevent the occurrence ofpremature death of nodes and prolong the network lifetime


Namely it indicates that it is suitable for the large-scaleWSNs Furthermore the deeper reason is that taking theaccumulated evidence of responsibility and availability as thedeterministic criterion for the cluster head election is reliableand promising On the other hand the scheme of Kruskal-based construction of the minimum aggregation tree ensuresthat the generated communication path is optimum which iseffectively against link losses

Additionally besides the LEACH algorithm has a littlefluctuation and ELBC has an abrupt change the lifetime ofthe network generated by the remainder three algorithmsshowsmore or less uptrendThe reason causing the changingsin LEACH algorithm is the unreasonable CHs electionmechanism For ELBC algorithm it has a relatively longerlifetimewhen the number of nodes is less than 200 its lifetimeeven reaches 621 within the size of 50 of nodes and thenthe network lifetime performs stably and remains at around390 which indicates that the ELBC algorithm only has abetter adaptability to the small-scale network The reasoncovered is primarily that the ELBC algorithm delivers datadirectly to the base station after the cluster formation whichmakes it easy to encounter the bottleneck problem of datatransmission such as data collision

Based on the above experimental analyses and compar-isons the effectiveness of the proposed EEMCR scheme isconvincing and its efficiency is promising

7 Conclusion and Future Works

In the light of network lifetime coverage preservationtransmission reliability and energy consumption a novelEEMCR scheme has been proposed In EEMCR scheme theguaranteeing capability of a link or a node of data transmis-sion is comprehensively considered The support of nodesthemselves and node-joint links to data transmission areabstracted as a propagation of responsibility and availabilitythat is the accumulated evidence The responsibility andavailability integrate considerations on the residual energyof nodes CH location energy consumption on the selectedcommunication path and distance between nodes The pre-sented cluster head rotation mechanism makes a globallyreasonable CH distribution proper size of clustering anduniform allocation of energy consumption On the otherhand backbone construction algorithm is employed to buildtheminimumaggregation tree within the candidate CHs andthe generatedminimum aggregation tree provides an energy-efficientmultihop routing to guarantee the optimal and short-est path between the CHs to the base station This schemecan effectively improve coverage preservation save energyand even degrade the link redundancy Although the EEMCRscheme inherits the infrastructure of LEACH algorithmthe theoretical analysis and empirical results demonstratethe promising performance over the LEACH algorithmsRegarding emergence of the first death node equilibriumof energy assignment rationality of CH distribution andconvergence rate of the proposed algorithm our scheme ispromising The proposed scheme is applicable for the large-scale WSNs In the future works the performance of ourscheme could be evaluated when embedded in the design

of new routing protocols Beyond that various kinds ofprotocols design and topology construction methods basedon the presented innovations for different layer of WSNswould be studied

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

This work is supported by the Future Research ProjectsFunds for the Science and Technology Department of JiangsuProvince (Grant no BY2013015-23) and the FundamentalResearch Funds for the Ministry of Education (Grant noJUSRP211A 41)

References

[1] O Durmaz Incel A Ghosh B Krishnamachari and K Chin-talapudi ldquoFast data collection in tree-based wireless sensornetworksrdquo IEEE Transactions on Mobile Computing vol 11 no1 pp 86ndash99 2012

[2] O Gnawali R Fonseca K Jamieson M Kazandjieva D Mossand P Levis ldquoCTP an efficient robust and reliable collectiontree protocol for wireless sensor networksrdquo ACM Transactionson Sensor Networks vol 10 no 1 article 16 2013

[3] Y Jin L Wang Y Kim and X Yang ldquoEEMC An energy-efficient multi-level clustering algorithm for large-scale wirelesssensor networksrdquo Computer Networks vol 52 no 3 pp 542ndash562 2008

[4] A P Renold and S Chandrakala ldquoSurvey on state scheduling-based topology control in unattended wireless sensor net-worksrdquo Computers and Electrical Engineering vol 56 pp 334ndash349 2016

[5] Z X Liu Q C Zheng L Xue and X P Guan ldquoA distributedenergy-efficient clustering algorithmwith improved coverage inwireless sensor networksrdquoFutureGenerationComputer Systemsvol 28 no 5 pp 780ndash790 2012

[6] S Soro and W B Heinzelman ldquoCluster head election tech-niques for coverage preservation in wireless sensor networksrdquoAd Hoc Networks vol 7 no 5 pp 955ndash972 2009

[7] C-H Lung and C Zhou ldquoUsing hierarchical agglomerativeclustering in wireless sensor networks an energy-efficient andflexible approachrdquo Ad Hoc Networks vol 8 no 3 pp 328ndash3442010

[8] W R Heinzelman A Chandrakasan and H Balakrish-nan ldquoEnergy-efficient communication protocol for wirelessmicrosensor networksrdquo in Proceedings of the 33rd AnnualHawaii International Conference on System Siences (HICSS rsquo00)vol 2 IEEE January 2000

[9] M Cobos F Antonacci A Alexandridis A Mouchtaris andB Lee ldquoA Survey of Sound Source Localization Methods inWireless Acoustic Sensor NetworksrdquoWireless Communicationsand Mobile Computing vol 2017 pp 1ndash24 2017

[10] O Younis and S Fahmy ldquoDistributed clustering in ad-hocsensor networks a hybrid energy-efficient approachrdquo in Pro-ceedings of the 23rd Annual Joint Conference of the IEEEComputer and Communications Societies ( INFOCOM rsquo04) vol1 pp 629ndash640 March 2004


[11] B Wang H B Lim D Ma and D Yang ldquoA coverage-awareclustering protocol for wireless sensor networksrdquo in Proceedingsof the 2010 6th International Conference on Mobile Ad-hoc andSensor Networks MSN 2010 pp 85ndash90 China December 2010

[12] L M C Arboleda and N Nasser ldquoComparison of clusteringalgorithms and protocols for wireless sensor networksrdquo inProceedings of the IEEE Canadian Conference on Electrical andComputer Engineering (CCECE rsquo06) pp 1787ndash1792 OttawaCanada May 2006

[13] C Li M Ye G Chen and J Wu ldquoAn energy-efficient unequalclusteringmechanism for wireless sensor networksrdquo in Proceed-ings of the 2nd IEEE International Conference on Mobile Ad-hocand Sensor Systems (MASS rsquo05) pp 597ndash604 Washington DCUSA November 2005

[14] E Leao C Montez R Moraes P Portugal and F VasquesldquoAlternative Path Communication in Wide-Scale Cluster-TreeWireless Sensor Networks Using Inactive Periodsrdquo Sensors vol17 no 5 p 1049 2017

[15] L-J Chen M Liu D-X Chen and L Xie ldquoTopology evolutionof wireless sensor networks among cluster heads by randomwalkersrdquo Jisuanji XuebaoChinese Journal of Computers vol 32no 1 pp 69ndash76 2009

[16] Y-H Zhu W-D Wu V C M Leung and L-H YangldquoEnergy-efficient tree-based message ferrying routing schemesfor wireless sensor networksrdquo in Proceedings of the 3rd Interna-tional Conference on Communications and Networking in ChinaChinaCom 2008 pp 843ndash847 China August 2008

[17] Y Jiang and H Zhang ldquoBase station controlled intelligentclustering routing in wireless sensor networksrdquo in Advances inArtificial Intelligence vol 6657 of Lecture Notes in Comput Scipp 210ndash215 Springer Heidelberg 2011

[18] K X Cui and Z H Li ldquoClutering Network Topology ControlAlgorithm for EBAPCBased onEnergyrdquoComputer Engineeringvol 38 no 23 pp 104ndash108 2012

[19] P F Li ldquoEnergy-level Based Clustering Network Control Algo-rithmrdquo Computer Science vol 41 no 3 pp 96ndash99 2014

[20] X Yin ldquoAffinity propagation clustering algorithm based onnode competing strengthrdquo Computer Engineering and Applica-tions vol 51 no 8 pp 79ndash84 2015

[21] W B Heinzelman A P Chandrakasan and H Balakrish-nan ldquoAn application-specific protocol architecture for wirelessmicrosensor networksrdquo IEEE Transactions onWireless Commu-nications vol 1 no 4 pp 660ndash670 2002

[22] Y Tao Y Zhang and Y Ji ldquoFlow-balanced routing for multi-hop clustered wireless sensor networksrdquo Ad Hoc Networks vol11 no 1 pp 541ndash554 2013

[23] M S Levin and A A Zamkovoy ldquoMulticriteria Steiner treewith the cost of Steiner verticesrdquo Journal of CommunicationsTechnology and Electronics vol 56 no 12 pp 1527ndash1542 2011

[24] B J Frey and D Dueck ldquoClustering by passing messagesbetween data pointsrdquo American Association for the Advance-ment of Science Science vol 315 no 5814 pp 972ndash976 2007

[25] M Younis P Munshi G Gupta and S M Elsharkawy ldquoOnefficient clustering of wireless sensor networksrdquo in Proceedingsof the DSSNS 2006 2nd IEEE Workshop on Dependability andSecurity in Sensor Networks and Systems pp 78ndash87 USA April2006

[26] O Younis M Krunz and S Ramasubramanian ldquoNode clus-tering in wireless sensor networks recent developments anddeployment challengesrdquo IEEE Network vol 20 no 3 pp 20ndash25 2006

[27] W Zhou H-M Chen and X-F Zhang ldquoAn energy efficientstrong head clustering algorithm for wireless sensor networksrdquoin Proceedings of the 2007 International Conference on WirelessCommunications Networking and Mobile Computing WiCOM2007 pp 2584ndash2587 China September 2007

[28] A Chamam and S Pierre ldquoA distributed energy-efficientclustering protocol for wireless sensor networksrdquo Computersand Electrical Engineering vol 36 no 2 pp 303ndash312 2010

[29] M Khalily-Dermany M Shamsi and M J Nadjafi-Arani ldquoAconvex optimization model for topology control in network-coding-based-wireless-sensor networksrdquoAdHoc Networks vol59 pp 1ndash11 2017

[30] B Yin H Shi and Y Shang ldquoAn efficient algorithm forconstructing a connected dominating set in mobile ad hocnetworksrdquo Journal of Parallel and Distributed Computing vol71 no 1 pp 27ndash39 2011

[31] S Lindsey and C S Raghavendra ldquoPEGASIS power-efficientgathering in sensor information systemsrdquo in Proceedings of theIEEE Aerospace Conference vol 3 pp 1125ndash1130 Big Sky MontUSA March 2002

[32] L Keller E Atsan K Argyraki and C Fragouli ldquoSenseCodenetwork coding for reliable sensor networksrdquo ACM Transac-tions on Sensor Networks vol 9 no 2 article 25 2013

[33] M R Garey and D S Johnson Computers and IntractabilityA Guide to the Theory of NP-Completeness W H Freeman SanFrancisco Calif USA 1979

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of


International Journal of

RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal of

Volume 201

Submit your manuscripts athttpswwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



Sink

0 10 20 30 40 50 60 70 80 90 1000

102030405060708090

100

1

2

3

4

5 67

8 9

10

11

12

13

14

15 16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

4445

46

47

48

49

50

51

52

53

54

55

56

57

5859

60

61

62

63

6465

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

9697

98

99

100

Nodes

CHs

The accumulatedevidenceThe optimum multihopcommunication paths

Figure 1 A demo of the principles employed in the proposedscheme

the shortest path tree with minimum energy consumptionfor each CH to the sink node Namely the CHs couldautomatically form a number of multihop communicationpaths The collected data is continuously transmitted to thesink node via the cluster head-adjacent multihop routing thateffectively shared the overloading of different CHs Howeverthe deficiency of imbalanced distribution of CHs has not beentreated which makes the size of some clusters too large andit is easy to incur the fast energy consumption in some localclusters

In this paper we also address the LEACH-improvedscheme By studying the broadcast nature of data trans-mission in WSNs it is abstracted as a propagation ofresponsibility and availability of nodes The responsibilityand availability integrate sufficient considerations on variousnetwork factors including the residual energy of nodesdistance between nodes distance between the CHs andthe base station and energy loss on the node-joint linksEssentially the responsibility and availability express theaccumulated evidence of nodes to support high-qualitycommunication Namely the responsibility and availabilityrepresent the comprehensive capability of nodes against nodefailures and link losses Based on the above developmentan evidence-efficient CH rotation scheme and algorithmare proposed Furthermore in order to improve the datatransmission reliability between the candidate CHs to thebase station the well-known Kruskal algorithm [23] isemployed to generate the minimum aggregation tree that isa backbone construction algorithm is also developed As ahybridization of the presented subalgorithms an evidence-efficient multihop clustering routing (EEMCR) method isproposed An example of the working mechanism employedin the EEMCR method is demonstrated by several clustersin Figure 1 And more the experimental comparison andanalysis are also examined with the previous algorithms

As far as we know the main contribution of this paperhas at least the following points (1) the eligibility evidenceof a node as a CH is regarded as the sum of responsibilityand availability (eg the accumulated evidence) whichincreases its robustness against random node failures (2)the mechanism of clustering is in terms of the propagation

of responsibility and availability globally which reduces thenegative intercluster communication interference (3) theoptimummultihop communication paths are achieved by thebackbone construction algorithm for the CHs which ensuresthe node-joint link reliability against link failures (4) theassociated algorithms and methods are developed and havevalidated their promising performance

The rest of this paper is organized as follows The relatedworks are introduced in Section 2 In Section 3 we describethe system assumptions and communication models InSection 4 the mechanism and algorithm of CH rotationare examined and proposed We derive the framework ofthe EEMCR scheme and present the overview and thedetailed design of it in Section 5 Promising experimentresults are given in Section 6 and from the effectiveness andefficiency perspective some validations and comparisons areperformed which are followed by the concluding remarksand future works in Section 7

2 Related Works

21 General Clustering Routing Grouping nodes into clustershas been themost popular approach for supporting scalabilityin WSNs [4] Besides the well-known LEACH and LEACH-developed algorithms significant attention has been paid toclustering routingmechanism and algorithms yielding a largeamount of publications [3 5ndash7 10ndash13 17 18 20 24] Theseproposed clustering techniques usually include three phasesCH determination clustering and data transmission Theprevious research works address either one of the issues oroverall three phases fromdifferent perspectives and also theirimprovements and developments with respect to the existingresearches

Usually clustering is typically based on the energy reserveof sensor nodes and nodersquos proximity to the CH [25]Regardless of any of their improvements or developmentthe reduction of energy consumption prolonging the net-work lifetime and improving transmission reliability are thebasic goals in WSNs Energy-efficient clustering techniqueis applied to reduce energy consumption interference andmaintaining connectivity and coverage in WSNs Youniset al [26] proposed a Hybrid Energy-Efficient DistributedClustering (HEED) protocol HEED is a distributed cluster-ing protocol in which a CH election scheme is presentedwith the comprehensive treatment of the residual energyand intracluster communication cost HEED ensures a uni-form distribution of CHs and inter-CH connectivity by anadjustable probability of CH election But the mechanismthat the sensor doubles its probability to become CH duringa repetition phase is not reasonable enough Zhou et al[27] proposed an Energy-Efficient Strong Head clustering(EESH) In EESH nodes are promotedCHs according to theirrespective residual energies their respective degrees and thedistance to and the residual energy of their neighbors Forthat EESH evaluates a cost function for every sensor in thenetwork and iteratively elects the node having the greatestcost as CH This process terminates when all the sensorsin the network are connected to at least one CH Chamamand Pierre [28] proposed a distributed energy-efficient












Amplifiercircuit

Transmittingcircuit

Receivingcircuit

d







119864TX (119897 119889) = 119864TX elec (119897) + 119864TX amp (119897 119889)=


(1)


1198890 = radic 120576fs120576amp (2)








NEL = (119864remain

119864init )1119873live (4)



EC (119894 119895) = ET (119894 119895)119864remain119894

(5)








119895



) (7)





119903 (119894 119896) = 119904 (119894 119896) minusmax 119886 (119894 1198961015840) + 119904 (119894 1198961015840)(1198961015840 isin 1 2 119899 1198961015840 = 119896) (8)

119886 (119894 119896) = min0 119903 (119896 119896) +sum1198941015840

max 0 119903 (1198941015840 119896)(1198941015840 isin 1 2 119899 1198941015840 = 119894 1198941015840 = 119896)

(9)



119903new (119894 119896) = (1 minus 120582) 119903 (119894 119896) + 120582119903old (119894 119896) (10)

119886new (119894 119896) = (1 minus 120582) 119886 (119894 119896) + 120582119886old (119894 119896) (11)













































0 50 100 1500

102030405060708090

100


0102030405060708090

100


















0 50 100 150 200 250 300 350 4000

102030405060708090

100

Cycles

Num

ber o

f dea

th n

odes

LEACHEBAPCELBC

BM-ELBCEEMCR





0 50 100 150 200 250 3000

5

10

15

20

25

30

Cycles

LEACHEBAPCELBC

BM-ELBCEEMCR

The t

otal

resid

ual e

nerg

y (J

)





0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

7

8

ID of nodes

LEACHEBAPCELBC

BM-ELBCEEMCR

The a

vera

ge en

ergy

cons

umpt

ion

(J)

times10minus3




100 110 120 130 140 150 160 170 180 190 2000

50100150200250300350400450


LEACHEBAPCELBC

BM-ELBCEEMCR

deat

h no

deCy

cles u

ntil

the fi

rst


100 110 120 130 140 150 160 170 180 190 200100

200

300

400

500

600

700

800


LEACHEBAPCELBC

BM-ELBCEEMCR

Life

time





0 100 200 300 400 500 600 700 800 900100

200

300

400

500

600

700

800

Node number

Life

time

LEACHEBAPCELBC

BM-ELBCEEMCR














Acknowledgments


References



































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation




















Amplifiercircuit

Transmittingcircuit

Receivingcircuit

d







119864TX (119897 119889) = 119864TX elec (119897) + 119864TX amp (119897 119889)=


(1)


1198890 = radic 120576fs120576amp (2)








NEL = (119864remain

119864init )1119873live (4)



EC (119894 119895) = ET (119894 119895)119864remain119894

(5)








119895



) (7)





119903 (119894 119896) = 119904 (119894 119896) minusmax 119886 (119894 1198961015840) + 119904 (119894 1198961015840)(1198961015840 isin 1 2 119899 1198961015840 = 119896) (8)

119886 (119894 119896) = min0 119903 (119896 119896) +sum1198941015840

max 0 119903 (1198941015840 119896)(1198941015840 isin 1 2 119899 1198941015840 = 119894 1198941015840 = 119896)

(9)



119903new (119894 119896) = (1 minus 120582) 119903 (119894 119896) + 120582119903old (119894 119896) (10)

119886new (119894 119896) = (1 minus 120582) 119886 (119894 119896) + 120582119886old (119894 119896) (11)













































0 50 100 1500

102030405060708090

100


0102030405060708090

100


















0 50 100 150 200 250 300 350 4000

102030405060708090

100

Cycles

Num

ber o

f dea

th n

odes

LEACHEBAPCELBC

BM-ELBCEEMCR





0 50 100 150 200 250 3000

5

10

15

20

25

30

Cycles

LEACHEBAPCELBC

BM-ELBCEEMCR

The t

otal

resid

ual e

nerg

y (J

)





0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

7

8

ID of nodes

LEACHEBAPCELBC

BM-ELBCEEMCR

The a

vera

ge en

ergy

cons

umpt

ion

(J)

times10minus3




100 110 120 130 140 150 160 170 180 190 2000

50100150200250300350400450


LEACHEBAPCELBC

BM-ELBCEEMCR

deat

h no

deCy

cles u

ntil

the fi

rst


100 110 120 130 140 150 160 170 180 190 200100

200

300

400

500

600

700

800


LEACHEBAPCELBC

BM-ELBCEEMCR

Life

time





0 100 200 300 400 500 600 700 800 900100

200

300

400

500

600

700

800

Node number

Life

time

LEACHEBAPCELBC

BM-ELBCEEMCR














Acknowledgments


References



































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











119895



) (7)





119903 (119894 119896) = 119904 (119894 119896) minusmax 119886 (119894 1198961015840) + 119904 (119894 1198961015840)(1198961015840 isin 1 2 119899 1198961015840 = 119896) (8)

119886 (119894 119896) = min0 119903 (119896 119896) +sum1198941015840

max 0 119903 (1198941015840 119896)(1198941015840 isin 1 2 119899 1198941015840 = 119894 1198941015840 = 119896)

(9)



119903new (119894 119896) = (1 minus 120582) 119903 (119894 119896) + 120582119903old (119894 119896) (10)

119886new (119894 119896) = (1 minus 120582) 119886 (119894 119896) + 120582119886old (119894 119896) (11)













































0 50 100 1500

102030405060708090

100


0102030405060708090

100


















0 50 100 150 200 250 300 350 4000

102030405060708090

100

Cycles

Num

ber o

f dea

th n

odes

LEACHEBAPCELBC

BM-ELBCEEMCR





0 50 100 150 200 250 3000

5

10

15

20

25

30

Cycles

LEACHEBAPCELBC

BM-ELBCEEMCR

The t

otal

resid

ual e

nerg

y (J

)





0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

7

8

ID of nodes

LEACHEBAPCELBC

BM-ELBCEEMCR

The a

vera

ge en

ergy

cons

umpt

ion

(J)

times10minus3




100 110 120 130 140 150 160 170 180 190 2000

50100150200250300350400450


LEACHEBAPCELBC

BM-ELBCEEMCR

deat

h no

deCy

cles u

ntil

the fi

rst


100 110 120 130 140 150 160 170 180 190 200100

200

300

400

500

600

700

800


LEACHEBAPCELBC

BM-ELBCEEMCR

Life

time





0 100 200 300 400 500 600 700 800 900100

200

300

400

500

600

700

800

Node number

Life

time

LEACHEBAPCELBC

BM-ELBCEEMCR














Acknowledgments


References



































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation















































0 50 100 1500

102030405060708090

100


0102030405060708090

100


















0 50 100 150 200 250 300 350 4000

102030405060708090

100

Cycles

Num

ber o

f dea

th n

odes

LEACHEBAPCELBC

BM-ELBCEEMCR





0 50 100 150 200 250 3000

5

10

15

20

25

30

Cycles

LEACHEBAPCELBC

BM-ELBCEEMCR

The t

otal

resid

ual e

nerg

y (J

)





0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

7

8

ID of nodes

LEACHEBAPCELBC

BM-ELBCEEMCR

The a

vera

ge en

ergy

cons

umpt

ion

(J)

times10minus3




100 110 120 130 140 150 160 170 180 190 2000

50100150200250300350400450


LEACHEBAPCELBC

BM-ELBCEEMCR

deat

h no

deCy

cles u

ntil

the fi

rst


100 110 120 130 140 150 160 170 180 190 200100

200

300

400

500

600

700

800


LEACHEBAPCELBC

BM-ELBCEEMCR

Life

time





0 100 200 300 400 500 600 700 800 900100

200

300

400

500

600

700

800

Node number

Life

time

LEACHEBAPCELBC

BM-ELBCEEMCR














Acknowledgments


References



































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation















0 50 100 150 200 250 300 350 4000

102030405060708090

100

Cycles

Num

ber o

f dea

th n

odes

LEACHEBAPCELBC

BM-ELBCEEMCR





0 50 100 150 200 250 3000

5

10

15

20

25

30

Cycles

LEACHEBAPCELBC

BM-ELBCEEMCR

The t

otal

resid

ual e

nerg

y (J

)





0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

7

8

ID of nodes

LEACHEBAPCELBC

BM-ELBCEEMCR

The a

vera

ge en

ergy

cons

umpt

ion

(J)

times10minus3




100 110 120 130 140 150 160 170 180 190 2000

50100150200250300350400450


LEACHEBAPCELBC

BM-ELBCEEMCR

deat

h no

deCy

cles u

ntil

the fi

rst


100 110 120 130 140 150 160 170 180 190 200100

200

300

400

500

600

700

800


LEACHEBAPCELBC

BM-ELBCEEMCR

Life

time





0 100 200 300 400 500 600 700 800 900100

200

300

400

500

600

700

800

Node number

Life

time

LEACHEBAPCELBC

BM-ELBCEEMCR














Acknowledgments


References



































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0 50 100 150 200 250 3000

5

10

15

20

25

30

Cycles

LEACHEBAPCELBC

BM-ELBCEEMCR

The t

otal

resid

ual e

nerg

y (J

)





0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

7

8

ID of nodes

LEACHEBAPCELBC

BM-ELBCEEMCR

The a

vera

ge en

ergy

cons

umpt

ion

(J)

times10minus3




100 110 120 130 140 150 160 170 180 190 2000

50100150200250300350400450


LEACHEBAPCELBC

BM-ELBCEEMCR

deat

h no

deCy

cles u

ntil

the fi

rst


100 110 120 130 140 150 160 170 180 190 200100

200

300

400

500

600

700

800


LEACHEBAPCELBC

BM-ELBCEEMCR

Life

time





0 100 200 300 400 500 600 700 800 900100

200

300

400

500

600

700

800

Node number

Life

time

LEACHEBAPCELBC

BM-ELBCEEMCR














Acknowledgments


References



































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










100 110 120 130 140 150 160 170 180 190 2000

50100150200250300350400450


LEACHEBAPCELBC

BM-ELBCEEMCR

deat

h no

deCy

cles u

ntil

the fi

rst


100 110 120 130 140 150 160 170 180 190 200100

200

300

400

500

600

700

800


LEACHEBAPCELBC

BM-ELBCEEMCR

Life

time





0 100 200 300 400 500 600 700 800 900100

200

300

400

500

600

700

800

Node number

Life

time

LEACHEBAPCELBC

BM-ELBCEEMCR














Acknowledgments


References



































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation


















Acknowledgments


References



































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation

































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









evidence-efficient multihop clustering routing scheme for...

Documents