a center-based secure and stable clustering algorithm for...

Research ArticleA Center-Based Secure and Stable ClusteringAlgorithm for VANETs on Highways

Xiaolu Cheng1 and Baohua Huang 2

1Virginia Commonwealth University Richmond Virginia 23220 USA2Guangxi University Nanning Guangxi 530004 China

Correspondence should be addressed to Baohua Huang bhhuang66gxueducn

Received 15 October 2018 Revised 5 December 2018 Accepted 20 December 2018 Published 2 January 2019

Guest Editor Zaobo He

Copyright copy 2019 Xiaolu Cheng and Baohua Huang This is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

Currently communications in the vehicular ad hoc network (VANET) can be established via both Dedicated Short RangeCommunication (DSRC) and mobile cellular networks To make use of existing Long Term Evolution (LTE) network in datatransmissions many methods are proposed tomanage VANETs Grouping the vehicles into clusters and organizing the network byclusters are one of the most universal and most efficacious ways Since the high mobility of vehicles makes VANETs different fromother mobile ad hoc networks (MANETs) the previous cluster-basedmethods for MANETsmay have trouble for VANETs In thispaper we introduce a center-based clustering algorithm to help self-organized VANETs forming stable clusters and decrease thestatus change frequency of vehicles on highways and two metrics A novel Cluster Head (CH) selection algorithm is also proposedto reduce the impact of vehiclemotion differencesWe also introduce twometrics to improve the security of VANETs A simulationis conducted to compare our mechanism to some other mechanisms The results show that our mechanism obtains high stabilityand lower packet loss rate

1 Introduction

As a key component of Intelligent Transportation Systems(ITS) vehicular ad hoc network (VANET) has attractedplenty of researchers from different fields and massiveresearch efforts have been made

In VANETs there are two types of communications [1]VANETs enable both vehicle-to-vehicle (V2V) communica-tions and vehicle-to-infrastructure (V2I) communications InVANETs vehicles and the infrastructures such as RoadsideUnits (RSU) and application servers exchange informationfor navigation safe driving entertainment and so on

Generally communications in VANETs are roughly cat-egorized into two classes according to the adopted radiointerfaces One class of approaches is based on DedicatedShort Range Communication (DSRC) The other class isbased on existing cellular technology [2]

DSRC began to be used for V2V communication fromthe 90s It has a shortage in medium range which is about300 meters It is inadequate for large-scale deployment

[3] because its coverage radius is not large enough Withthe rapid improvement of mobile cellular networks someresearchers supposed to utilize the existing mobile cellu-lar infrastructures and technologies for communications inVANETs Mobile cellular networks provide wider and largercoverage while their delay is longer than DSRC for real-time information exchanges in local areas [4]Therefore bothDSRC and mobile cellular networks cannot fully meet theneeds of ITS As a result VANETs support communicationnot only via LTE but also via DSRC

To make use of existing mobile cellular networks fordata transmissions many methods are proposed to man-age VANETs However if VANETs are fully managed byinfrastructures low efficiency will be a big issue whilefully decentralized VANETs must create a lot of overheadTherefore VANETs usually combine some centralized partsand decentralized parts To decrease the overhead via DSRCchannels and the probability of LTE channel congestionVANETs are centralized by cellular-based connections and

HindawiWireless Communications and Mobile ComputingVolume 2019 Article ID 8415234 10 pageshttpsdoiorg10115520198415234

2 Wireless Communications and Mobile Computing

scheduling Meanwhile vehicles may also exchange messageswith their neighbors viaDSRCDividing vehicles into clustersis a common and reasonable approach for VANETs man-agement In a cluster-based framework vehicles are signedinto clusters The range of a cluster is smaller or equal tothe range of 80211p so that vehicles in the same clustercan exchange messages via DSRC A single eNodeB managesmany clusters around it Within a cluster at least one vehicleperforms as a Cluster Head (CH) to collect information ofall Cluster Members (CM) via DSRC and exchanges datawith the eNodeB via TLE This architecture decreases themanagement overhead while utilizing both DSRC and LTE

Compared to other MANETs nodes in VANETs havehigher mobility and higher speed Cluster reforming andCH changing must be much more frequent than othertypical MANETs To decrease the management overhead andincrease communication quality the clustering algorithm forVANETs should be able to form stable clusters To achieve thisgoal in this paperwe propose a stable clustering algorithm forVANETsWe propose a novel approach to form and maintainstable clusters for VANETs on highways to avoid continualcluster reforming A center-based clustering algorithm isused to locate the initial clustersrsquo centers In every clustera suitable CH is chosen by vehiclesrsquo position speed andmaximal acceleration A cluster maintenance algorithm isproposed to keep CMs in its CHrsquos transmission range

The rest of the paper is organized as follows The RelatedWork briefly reviews the current literature on clusteringalgorithms in VANETs The proposed scheme is detailed inthe Proposed SchemeThe simulation parameters simulationresults and analysis are shown in the Performance Evalua-tion In the Conclusion we state the conclusion

2 Related Work

In the literature clustering is the process to group vehicles inVANETS

Ref [5] proposes a method named LTE4V2X to orga-nize vehicular networks In the centralized vehicular net-works eNodeB manages vehicles in its coverage and dividesthem into clusters LTE4V2X protocol defines how the self-organized network works In LTE4V2X eNodeB creates clus-ters which contain the largest number of nodes circulating inthe same direction

Ref [6] extends LTE4V2X to increase information dis-semination efficiency It selects CH by the distance fromvehicles to eNodeB Although comparing to the originalapproach the complexity is lower and the LTE channel qual-ity is higher the power consumption of message exchangingis not optimized Nevertheless [6] states that the systemcan calculate the transmit power of DSRC channels by thedistance between vehicles so that the transmit power couldbe dynamically adjusted

Road condition affects the speed and direction of vehiclesFor example vehiclersquos speed is lower on the bumpy road thana smooth road Vehicle mobility is determined by humanbehavior Take a street connected megapolis and a village asan example In the morning most vehicles move from thevillage (home) to the megapolis (office) In the evening most

vehicles run following the reverse path Ref [7] quantifiestemporal locality similarity to measure the relation of twovehiclesrsquo mobility Then they utilize the relation of vehiclesrsquomovements to form stable clusters The locality can also beused for reducing energy consumption [8]

Ref [9] proposes a clustering approach to minimize thetotal power consumed by DSRC communications They usea weighted distance matrix to indicate power consumedbetween each pair of vehicles In this way the CH selectionproblem is formulated as a variant of the p-median problemin graph theory [10] In this approach the number of clusters119901 is determined first based on LTE coverage radius andDSRC coverage radiusThe119901 cluster zones are determined byvehicle number and 80211P coverage radius p Cluster Headsthat are closet to the eNodeB are selected Then the systemdynamically selects new CH to minimize the transmissionpower between CMs and CH based on weighted distance andthe p-median issue in graph theory Although this approachminimizes the power consumption within a single cluster thepower consumption of V2I communications has not beenconsidered The method to decide the zones is vague andcomplicated Moreover this approach is not suitable for thescenario that CMs not only send their information to CH butalso communicate among themselves

Ref [11] proposes a high-integrity file transfer schemefor VANETs on highways named Cluster-based File Transfer(CFT) scheme In this scheme CMs help their CH todownload file fragments and then transmit fragments tothe CH which requests the file Since the very high speedof vehicles on highways CFT is a good approach to helpthe vehicles download files which they have not enoughconnection time to download However CFT just considersthe bidirection environment In addition with CFT CHbroadcasts its request to its neighbors then neighbors thatreceive the invitation join the cluster and broadcast therequest to invite more vehicles to join the cluster untilthere are enough vehicles Therefore CFT may not able toapply in complicated environment and it may cause networkcongestions

Ref [12] proposes an evolutionary game theoretic (EGT)framework for clustering and CH selecting Their protocolis based on game theory They defined the net utility of aCH to select the CH which may achieve high throughput Acluster size is added in the utility function for CH to optimizethe size of a cluster Ref [13] proposes an intelligent naiveBayesian probabilistic estimation practice (ANTSC) methodThis method is based on the traffic flow To increase thestability of cluster a CH must be in the lane having theheaviest traffic flow Naive Bayes algorithm is used to selectthe CH which may make the cluster most stable However[12] just compared the EGT clustering with one clusteringalgorithm proposed in 2010 Both [12 13] did not considerthe security of the network

3 Proposed Scheme

31 Overview and Assumption Clustering algorithm groupsa set of unlabeled nodes into clusters In cluster-basedVANETs all vehicles send their information to eNodeB

Wireless Communications and Mobile Computing 3

eNodeB

CH

CM

LTE

802 11P

Figure 1 Communications within one cluster

Then eNodeB manages the vehicles by clusters A CH acts asa messenger to help eNodeB and CMs exchange information

We assume all vehicles are able to communicate via bothLTE and DSRC The size of cluster is smaller or equal tothe range of 80211p so that vehicles in the same clustercan exchange messages via DSRC DSRC coverage radius isabout 300 meters LTE coverage radius is about 1 kilometerTherefore a single eNodeB manages many clusters around itWithin a cluster a vehicle acts as a CH to collect informationof all CMs via 80211p and exchanges datawith the eNodeB viaTLE Figure 1 is a simplified view of a cluster-based vehicularnetwork

For cluster in this paper we have some assumptions

(1) All vehicles have both LTE and 80211p interfaces(2) All vehicles are equipped with Global Positioning

System (GPS) devises So they have accurate geolo-cations

(3) All vehicles know their destination speed and maxi-mal acceleration

Based on the assumptions we propose a center detectionbased clustering algorithm We group the vehicles in theregion where the density of vehicles is higher than other areas

into clusters with the help of blob detection method or animproved high-degree algorithm Some parameters such asspeed and acceleration are added to the CH selection metricto make the cluster stabler and decrease the CH reselectionfrequency

32 Cluster Formation In our proposed algorithm in theinitialization stage of cluster formation vehicles send beaconmessages to the eNodeB The beacon message of one vehiclecontains the vehiclersquos ID k current position (119909119896 119910119896) currentspeed V119896 maximal acceleration 119886119896 and direction type 119905119896

Direction type is decided by the angle from the currentposition to the destination For vehicle k whose destinationposition is (1199091015840119896 1199101015840119896) the direction angle 120579119896 is

120579119896 = tanminus11199101015840119896minus 119910119896

1199091015840119896 minus 119909119896 (1)

When 120579119896 isin [0∘ 90∘) 119905119896 = 1 When 120579119896 isin [90∘ 180∘) 119905119896 = 2When 120579119896 isin [180∘ 270∘) 119905119896 = 3 When 120579119896 isin [270∘ 360∘) 119905119896 =4 Vehicles that have different 119905 are managed respectively

The clustering algorithm is described in Algorithm 1After receiving the beacon messages the system analyzes

vehiclesrsquo position information and detects the centers of theranges where the vehicle density is higher than in otherareas If the vehicle quantity or the vehicle density is notvery large an improvedHighest-DegreeAlgorithm is appliedSeveral vehicles which have more neighbors in their transmitrange are detected We improve the original Highest-DegreeAlgorithm to make sure the distance between any twovehicles we detected is larger than the DRSC range Thepositions of detected vehicles will be the centers we usein the clustering algorithm Otherwise when the vehiclequantity and the vehicle density are very large to decrease thecomputing complexity and analyze time the system drawsdots on the map to indicate vehicles Then we can carry outthe blob detection The blob detection is able to detect theregions where the gray pixel value is greaterThus we can usethe blob detection algorithm eg [14] to detect the centersof regions on the map where dot density is higher

All vehicles whose distances to the center are not largerthan the range of DSRC are labeled as one cluster Thenthe system selects one nearest intersection for every centeramong all intersections that meet the following conditions

(1) The distance from it to the points in 119875 is not smallerthan the range of DSRC

(2) The intersection is not in any clusterrsquos region

Vehicles near those selected intersections are grouped intoclusters Then eNodeB uses the same way to select inter-sections near the selected intersections and groups vehiclesAfter iterations ungrouped vehicles are grouped into clustersThe distance between two vehicles in the same cluster isnot larger than the range of DRSC To further decreasecomputing complexity in line 8 of clustering algorithm avehicle or infrastructure located in the center or intersectioncan broadcast a request to invite neighbors to join the clusterIn line 37 the chosen vehicle 119890 can broadcast an invitationinstead of calculating distance by the system


Input Vehicle set VOutput Initial clusters

1 Initialize center set C = 1206012 Locate the centers and add them into 1198623 Initialize point set 119875= 1198624 while 119875 = 120601 do5 foreach point 119901 in 119875 do6 Initialize node set clusterp = 1206017 foreach vehicle e in 119881 do8 if dep le 119877 then9 Add 119890 into clusterp10 Remove 119890 from 11988111 end12 end13 if clusterp = 120601 then14 Call Algorithm 215 Return set clusterp16 Estimate Stc of the CH 119888 of clusterp17 if Stc is remarkable high then18 Check all nodes in clusterp to detect attacker19 end20 else21 119888 check 119878

119886value of each CM

22 if Sam is remarkable high then23 Report to the server24 end25 end26 Add the intersection nearest to 119901 which meets the conditions into 11987527 end28 Remove 119901 from 11987529 end30 end31 while V = 120601 do32 foreach point c in C do33 Select an element e in V nearest to c34 Initialize set clustere= 11989035 Remove e from V 36 Set e as CH37 foreach vehicle v in 119881 do38 if 119889

119890V le 119877 then39 Add V into clustere40 Remove V from 11988141 end42 end43 Return clustere44 Estimate Stcof the CH e of clustere45 if Stc is remarkable high then46 Check all nodes in clustere to detect attacker47 end48 else49 c check 119878


50 if Sam is remarkable high then51 Report to the server52 end53 end54 end55 end

Algorithm 1 Clustering algorithm


33 Cluster Head Selection Compared to other MANETsVANETs have lower stability because of the high mobilityof vehicles Although we divide the vehicles with the help ofdirection vector 997888rarrV 119896 the stability of clusters cannot be guar-anteed To select an appropriate CH which can increase thecluster lifetime and decrease the CH reselecting frequency arelative mobility metric119872 is introduced for CH election

For a vehicle k which is in the cluster clusteri the positiondifference between it and all other 119873 vehicles in the samecluster clusteri is

119863119896 =119873

sum119899=1

radic(119909119896 minus 119909119899)2 + (119910119896 minus 119910119899)2 (2)

The speed difference between 119896 and all other119873 vehicles in thesame cluster is

119881119896 =119873

sum119899=1

1003816100381610038161003816V119896 minus V1198991003816100381610038161003816 (3)

The maximal acceleration difference between 119896 and all other119873 vehicles in the same cluster is

119860119896 =119873

sum119899=1

1003816100381610038161003816119886119896 minus 1198861198991003816100381610038161003816 (4)

The relative mobility metric119872 is

M119896 = 120572 119863119896max 119863119899 | forall119899 isin 119862119894 + 120573 119881119896

max 119881119899 | forall119899 isin 119862119894+ 120574 119860119896

max 119860119899 | forall119899 isin 119862119894 (5)

where120572 120573 and 120574 are weighted coefficients 120572+120573+120574 = 1Theycan be adjusted to fit the different traffic conditions Whenthe traffic condition is good and all vehicles are driving ata similar speed the distance between vehicles has a greatereffect Thus the value of 120572 should be higher than the othertwo When vehicles are driving at high speed the value of 120573is higher than the other two When the vehicles enter an areawhich speed limit changes continually a higher 120574 should beconsidered

The relative mobility metric119872 evaluates the relative posi-tion speed and maximal acceleration differences betweenone vehicle and all other vehicles in the same cluster Asmaller 119872 indicates the vehicle has lower relative mobilitythan other vehicles in this cluster Algorithm 2 explains theprocess of Cluster Head selection All clusters formed withthe help of centers and intersections use Algorithm 2 to selectCH As a CH the vehiclersquos relative mobility metric is smallerthan any CMsThat means the motion mode of CH is similarto the whole cluster

34 Cluster Maintenance and Reforming The unpredictabil-ity andmobility of trafficmake the cluster lifetime temporaryIt is infeasible to reform clusters in real time or very fre-quently To minimize the frequency and overhead of cluster

Input Vehicles in one clusterOutput Cluster head o of the corresponding cluster

1 SetM119898119894119899

= +infin2 foreach vehicle 119896 do3 Calculate the relative mobility metricM

119896

4 if119872119896 lt119872119898119894119899 then5 M

119898119894119899= M119896

6 119900 = 1198967 end8 end9 return o

Algorithm 2 Cluster head selection algorithm

reforming we propose a cluster maintenance algorithmAlgorithm 3 explains the cluster maintenance process

(1) No Connections between CH and CM When a CH cannotconnect to a CM the CH will delete the CM from its recordand notice eNodeBWhen a CM cannot reach its CH the CMwill check the signal it received via DSRC and join the clusterwhose signal of CH is strongest If the CM cannot receive amessage strong enough it will notice eNodeB via LTE andbecome a CH

(2) No Connections between eNodeB and CH When eNodeBnotices it has lost connection to a CH it recalls Cluster HeadSelection Algorithm and a new vehicle will be CH of thatcluster instead of the leaving vehicle

(3) A Vehicle Joins the Network When a vehicle comesinto the network it first tries to join the nearest cluster bybroadcasting a CH request via DSRC If it fails it will senda message to eNodeB eNodeB will help the vehicle to join acluster or to be a CH and form a new cluster by itself

(4) Two Clusters Are Too Close With the movement of thevehicles two clusters may be very close When the distancebetween two CHs is shorter than 119877 for a period Δ119905 thetwo clusters are merged into one cluster The Cluster HeadSelected Algorithm is recalled A new CH for the newcluster is selected Then all vehicles which are out of thetransmission range of the new CH leave this cluster andcheck the invitation signal they have received via DSRC andjoin the clusterwhose signal of CH is the strongest If a vehicledoes not find a cluster to join in it notices eNodeB via LTEand becomes a CH

35 Security Mechanism To further improve the VANETssecurity and availability a novel security mechanism isproposed to detect malicious nodes

In clustered networks the availability and security ofCHs are incredibly crucial CHs help the servers to collectand transmit messages to CMs If an attacker wants theaccess to other vehiclesrsquo private information it should actsas a CH The most common and most executable methodfor an attacker to be selected as a CH is launching a


Input Initial clusters and vehicle set VOutput New clusters

1 if the eNodeB can not reach a CH then2 Call Cluster Head Selection Algorithm3 end4 if the CH can not reach a CM then5 Reomve the CM6 Notice eNodeB7 end8 if the distance between two CHs le 119877 for a period Δ119905 then9 Merge the two clusters into one cluster10 Call the Cluster Head Selected Algorithm11 Check 119878

119905values of new CHs

12 New CHs check 119878119886values of their CMs

13 end14 if a CM can not reach the CH then15 if it can receive a signal from CHs then16 Join the cluster whose signal of CH is strongest17 CH check its 119878

119886value

18 end19 else20 Notice eNodeB21 The node performs as a CH22 end23 end

Algorithm 3 Cluster maintenance algorithm

Sybil attack In a Sybil attack the vehicle controlled by amalicious attacker presents multiple identities (vehicles) andall of the vehicles have similar directions positions speedsand maximal acceleration Hence these vehicles must havehigher relative mobility metrics and higher probabilities tobe selected as CH

To protect the CMsrsquo privacy we introduce a trajectorysimilarity metric 119878119905 to defend Sybil attacks For a clustercontains 119873 nodes the trajectory similarity metric of its CH119888 is

119878119905119888 = 1119873 minus 1

119873minus1

sum119894=1

( Δ119905119894119897119894119891119890119905119894119898119890119894) (6)

where Δ119905119894 is the duration of both 119888 and node 119894 that belongto the same cluster and lifetimei is the lifetime of 119894 in thisnetwork Every time a CH is selected the server estimatesits trajectory similarity metric If one CH has a remarkablehigher trajectory similarity metric the server will check allnodes in the cluster to detect the malicious attacker

A denial-of-service attack (DoS attack) is another com-mon attack in VANETs In DoS attack the attacker floodsthe CH or server to make the network services unavailableThe connections of authenticated vehicles to the networkare temporarily broken Therefore the legitimate requests ofserver and authenticated vehicles cannot be actioned

To protect the network availability we introduce anactivity similarity metric 119878119886 to detect DoS attacks For a

vehicle 119898 that belongs to a cluster containing 119873 nodes theactivity similarity metric is

119878119886119898 = sum119873minus1119894=1

119901119894 minus 119901119898sum119873minus1119894=1

119901119894 sdot 119873119873 minus 1 (7)

where 119901119894 is the number of requests between vehicle 119894 andCH 119888 during a period of time Δ119905 119901119898 is the number ofrequests between vehicle119898 andCH 119888 in the same timeperiodThe higher the activity similarity metric the greater theproportion of requests of vehicle 119898 in all the requests of thiscluster If one CM has a remarkable higher similarity metrictheCHwill regard it as aDoS attacker and report to the serverMeanwhile the sever can also use activity similarity metricsto detect DoS attack from CHs

4 Performance Evaluation

41 Simulation Parameter We perform the simulation withthe help of Veins LTE Veins LTE is a simulator developed onVeins [15] which is an open source framework for simulationof vehicular networks based on both IEEE 80211p and LTEIt integrates a network simulator named OMNeT++ anda traffic simulator named Simulation of Urban MObility(SUMO) [16]

In our experiment vehicles run on a real map of Wash-ington DC USA obtained from OpenStreetMap [17] Weextract the data of highways in the center ofWashington DCThe total length of road is 3038 km The total lane length is


9009km Every vehicle has random source and destinationedge The route from the starting point to the destinationis the shortest path found by Dijkstrarsquos algorithm [18] Themaximal acceleration ability of vehicles we have used is 26m1199042Themaximal deceleration ability of vehicles is 45m1199042The vehiclersquos maximum velocity is 5555ms

We compare our proposed clustering algorithm Center-Based Stable Clustering Algorithm (CBSC) with a K-Means-Based method (KMB) and SCalE algorithm [19] K-meansalgorithm [20] is commonly used in VANETs for clusteringeg [21ndash23] In KMB method we divide the vehicles intotwo parts by the angle of the vehicles and perform KMB onthem respectively The cluster maintenance algorithm KMBis proposed in [24] The predefined threshold ΔV119905ℎ is 5 msIn the simulations all vehiclesrsquo movement information isresent to eNodeB for cluster status update in every 10 secondseNodeB needs exchange data with vesicles every 3 secondsThe simulation time is 503 seconds

42 Results and Analysis The goal of this paper is to proposea stable clustering algorithm for VANETs To check whethera clustering algorithm can solve the high mobility of vehicleson the highways the cluster stability should be evaluatedThe metrics we use to show the performance of clusteringalgorithm are as follows

(1) Average CH Lifetime The CH lifetime is the periodfrom the state when the vehicle is a CH to the statewhen it is not a CH (ie being a CM or leaving thesystem) When a CH ends its lifetime a new CH iselected or the cluster is dissolved

(2) Average CM Lifetime CM lifetime represents theduration at which a CM stays in the same clusterThe average CM lifetime is the average length of allvehiclesrsquo CM lifetime It is another important metricto evaluate the stability of clusters

(3) Average Number of Reaffiliation Times per VehicleThe average number of reaffiliation times per vehiclerepresents the average number of times a vehiclechanges the cluster it belongs to during the simulationtime

(4) Packet Loss Rate Packet loss rate is the percentage ofpackets lost with respect to packets sent

In the experimentation we compare the four metrics of thethree methods with different vehicle numbers transmissionranges or highway speed limits Figures 2 4 6 and 8 show theresults obtained with the variety of total vehicle number (N)and the variety of transmission range (R) when the highwayspeed limit (v) is 100 kmh Figures 3 5 7 and 9 show theresults obtained with the variety of transmission range (R)and the variety of highway speed limit (v) when the totalvehicle number (N) is 300

Figures 2 and 3 represent the average CH lifetime forthe three methods Those figures show that the CHs underKMB have a marked shorter lifetime Although our CBSChas a higher value than SCalE a few times in generalSCalE performs slightly better than CBSC on the average CHlifetime

N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)

Figure 2 Average CH lifetime versus N and R

R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)

Figure 3 Average CH lifetime versus R and v

The averageCM lifetime values produced byKMB SCalEand the CBSC methods are shown in Figures 4 and 5From those two figures we can see that the average CMlifetime produced by CBSC ismuch longer than the other twomethods SCalE has the worst performance on the averageCM lifetime

Figures 6 and 7 show the average number of reaffiliationtimes per vehicle obtained in 503 seconds Obviously com-paring to other two algorithms vehicles with SCalE changestatus much more frequently The data on the two figuresshows CBSC not only produces a lower cluster status changefrequency than KMB produces its superiority but also isbigger with the increase in highway speed limit

The results of simulation illustrate that clusters underCBSC are the stablest in the three algorithms They havethe longest average CM lifetime and lowest average numberof reaffiliation times per vehicle Although SCalE performsslightly better than CBSC on the CH lifetime experiment itproduces a much shorter average CM lifetime Besides thenumber of CMs is much larger than the CHs in one system


N=1

50

R=20

0N

=150

R=

300

N=1

50

R=40

0N

=300

R=

200

N=3

00

R=30

0N

=300

R=

400

N=4

50

R=20

0N

=450

R=

300

N=4

50

R=40

0

KMBSCalECBSC

0102030405060708090

100

Aver

age C

M L

ifetim

e (s)

Figure 4 Average CM lifetime versus N and R

R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0102030405060708090

Aver

age C

M L

ifetim

e (s)


0123456789

10

N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC

Figure 6 Average number of reaffiliation times per vehicle versusN and R

0123456789

10

R=20

0v=

70

R=20

0v=

100

R=20

0v=

130

R=30

0v=

70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70

R=40

0v=

100

R=40

0v=

130

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC

Figure 7 Average number of reaffiliation times per vehicle versus Rand v

N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003Pa

cket

Los

s Rat

e (

)

Figure 8 Packet loss rate versus N and R

R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003

Pack

et L

oss R

ate (

)

Figure 9 Packet loss rate versus R and v


Therefore we consider that CBSC has higher stability thanSCalE

The basic function of VANETs is supporting commu-nication between separated vehicles and infrastructures Totest the performance of data dissemination in VANETs wedo experiment on packet loss rate with different methodsPacket loss means a packet fails to arrive at its destinationA high packet loss rate decreases the data disseminationefficiency and may cause network congestion Therefore anefficient data dissemination mechanism should have a lowpacket loss rate In our experiment all vehicles exchangedata with eNodeB every three seconds That means in everythree seconds eNodeB sends data to all vehicles once andeach vehicle sends data to eNodeB once Like the scene wedescribed in the previous section eNodeB communicateswith the nodes in its record via CHs and vehicles whichare CMs send data to their CHs first Figures 8 and 9 showthe results of packet loss rate With the increase in vehiclevelocity or the transmission range the packet loss ratesobtained by all the three mechanisms decrease But CBSCgets lower packet loss rate while KMB performs the worstwhen the amount of vehicle is larger That insinuates CBSChas a good ability to handle a considerable amount of dataIn the experiment CBSC always obtains lowest packet lossrate Since the interval between cluster status updates is 10seconds we can know that the probability of CM leaving itsCH between two cluster status updates in CBSC is lower thanthe other two algorithms We can consider that the proposedrelative mobility metric 119872 and CH selection algorithm ofCBSC do reduce the impact of vehicle mobility on clusterstability

5 Conclusion

To decrease the management overhead and increase thequality of communications we try to make the clusters inVANETs as stable as possible while keeping the networkperformance acceptable In this paper we propose a stableclustering algorithm for VANETs on highways which utilizesdirection vector the centers of vehicle denser areas andintersections to group less quantity of more stable clustersTo reduce the impact of vehicle type and driversrsquo drivinghabits we propose a novel CH selection algorithm and clus-ter maintenance algorithm which use the relative mobilitymetric to reduce the influence of vehiclersquos distance velocityand maximal acceleration To protect the vehiclesrsquo privacyand the network availability we introduce two mechanismsto detect malicious attacker In the simulation experimentour algorithmrsquos performance ranks up against the other twoalgorithms (KMB and SCalE) on both stability and packagedelivery rate In the future we would like to further improvethe algorithm for the complex urban environment

Data Availability

No data were used to support this study

Conflicts of Interest

The authors declare that they have no conflicts of interest

References

[1] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[2] J Luo and J P Hubaux ldquoA survey of inter-vehicle communica-tionrdquo Tech Rep 2004

[3] Y J Li ldquoAn overview of the dsrcwave technologyrdquo in Proceed-ings of the International Conference on Heterogeneous Network-ing for Quality Reliability Security and Robustness pp 544ndash558Springer 2010

[4] K Zheng Q Zheng P Chatzimisios W Xiang and Y ZhouldquoHeterogeneous vehicular networking a survey on architecturechallenges and solutionsrdquo IEEE Communications Surveys ampTutorials vol 17 no 4 pp 2377ndash2396 2015

[5] G Remy S-M Senouci F Jan and Y Gourhant ldquoLTE4V2XLTE for a centralized VANET organizationrdquo in Proceedings ofthe 54th Annual IEEE Global Telecommunications ConferenceldquoEnergizing Global Communicationsrdquo GLOBECOM 2011 pp 1ndash6 Houston TX USA December 2011

[6] AMemedi F Hagenauer F Dressler and C Sommer ldquoCluster-based transmit power control in heterogeneous vehicularnetworksrdquo in Proceedings of the IEEE Vehicular NetworkingConference VNC 2015 pp 60ndash63 December 2015

[7] Y LiM Zhao andWWang ldquoIntermittently connected vehicle-to-vehicle networks detection and analysisrdquo in Proceedings ofthe 54th Annual IEEE Global Telecommunications Conference(GLOBECOM rsquo11) pp 1ndash6 December 2011

[8] L Zhang Y Deng W Zhu J Zhou and F Wang ldquoSkewlyreplicating hot data to construct a power-efficient storageclusterrdquo Journal of Network and Computer Applications vol 50pp 168ndash179 2015

[9] P Dong X Du J Sun and H Zhang ldquoEnergy-efficient clustermanagement in heterogeneous vehicular networksrdquo in Pro-ceedings of the IEEE Conference on Computer CommunicationsWorkshops (INFOCOMWKSHPS) pp 644ndash649 San FranciscoCalif USA April 2016

[10] S L Hakimi ldquoOptimum distribution of switching centers ina communication network and some related graph theoreticproblemsrdquo Operations Research vol 13 pp 462ndash475 1965

[11] Q Luo C Li Q Ye T H Luan L Zhu and X Han ldquoCFT ACluster-based File Transfer Scheme for highway VANETsrdquo inProceedings of the ICC 2017 - IEEE International Conference onCommunications pp 1ndash6 Paris France May 2017

[12] A A Khan M Abolhasan and W Ni ldquoAn Evolutionary GameTheoretic Approach for Stable and Optimized Clustering inVANETsrdquo IEEE Transactions on Vehicular Technology vol 67no 5 pp 4501ndash4513 2018

[13] A Mehmood A Khanan A H H M Mohamed S MahfoozH Song and S Abdullah ldquoANTSC An Intelligent Naıve Baye-sian Probabilistic Estimation Practice for Traffic Flow to FormStable Clustering inVANETrdquo IEEEAccess vol 6 pp 4452ndash44612017

[14] T Lindeberg ldquoFeature detectionwith automatic scale selectionrdquoInternational Journal of Computer Vision vol 30 no 2 pp 79ndash116 1998

[15] C Sommer D Eckhoff R German and F Dressler ldquoAcomputationally inexpensive empirical model of IEEE 80211pradio shadowing in urban environmentsrdquo in Proceedings of the8th International Conference on Wireless On-Demand NetworkSystems and Services (WONS rsquo11) pp 84ndash90 January 2011


[16] M Behrisch L Bieker J Erdmann and D KrajzewiczldquoSumondashsimulation of urban mobility an overviewrdquo in Proceed-ings of the SIMUL 2011 The Third International Conference onAdvances in System Simulation ThinkMind 2011

[17] (OSMF) O F ldquoOpenstreetmaprdquo httpwwwopenstreetmaporgmap=14389004-770282

[18] E W Dijkstra ldquoA note on two problems in connexion withgraphsrdquoNumerische Mathematik vol 1 no 1 pp 269ndash271 1959

[19] G V Rossi Z Fan W H Chin and K K Leung ldquoStableclustering for Ad-Hoc vehicle networkingrdquo in Proceedings ofthe 2017 IEEE Wireless Communications and Networking Con-ference WCNC 2017 pp 1ndash6 March 2017

[20] J MacQueen ldquoSome methods for classification and analysis ofmultivariate observationsrdquo in Proceedings of the fifth Berkeleysymposium on mathematical statistics and probability vol 1 no14 pp 281ndash297 Oakland CA USA 1967

[21] E Ben Hamida and M A Javed ldquoChannel-aware ECDSAsignature verification of basic safety messages with K-meansclustering in VANETsrdquo in Proceedings of the 30th IEEE Inter-national Conference on Advanced Information Networking andApplications AINA 2016 pp 603ndash610 March 2016

[22] Q Zhang M Almulla Y Ren and A Boukerche ldquoAn efficientcertificate revocation validation scheme with k-means cluster-ing for vehicular ad hoc networksrdquo in Proceedings of the 2012IEEE Symposium on Computers and Communications (ISCC)pp 000862ndash000867 Cappadocia Turkey July 2012

[23] R Chai X Ge and Q Chen ldquoAdaptive K-Harmonic Meansclustering algorithm for VANETsrdquo in Proceedings of the 14thInternational Symposium on Communications and InformationTechnologies ISCIT 2014 pp 233ndash237 IEEE September 2014

[24] Z Y Rawashdeh and S M Mahmud ldquoA novel algorithm toform stable clusters in vehicular ad hoc networks on highwaysrdquoEURASIP Journal onWireless Communications andNetworkingvol 2012 no 1 p 15 2012

International Journal of

AerospaceEngineeringHindawiwwwhindawicom Volume 2018

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018


Active and Passive Electronic Components

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawiwwwhindawicom

Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Control Scienceand Engineering

Journal of



Journal ofEngineeringVolume 2018

SensorsJournal of



RotatingMachinery


Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation


Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

Submit your manuscripts atwwwhindawicom


scheduling Meanwhile vehicles may also exchange messageswith their neighbors viaDSRCDividing vehicles into clustersis a common and reasonable approach for VANETs man-agement In a cluster-based framework vehicles are signedinto clusters The range of a cluster is smaller or equal tothe range of 80211p so that vehicles in the same clustercan exchange messages via DSRC A single eNodeB managesmany clusters around it Within a cluster at least one vehicleperforms as a Cluster Head (CH) to collect information ofall Cluster Members (CM) via DSRC and exchanges datawith the eNodeB via TLE This architecture decreases themanagement overhead while utilizing both DSRC and LTE

Compared to other MANETs nodes in VANETs havehigher mobility and higher speed Cluster reforming andCH changing must be much more frequent than othertypical MANETs To decrease the management overhead andincrease communication quality the clustering algorithm forVANETs should be able to form stable clusters To achieve thisgoal in this paperwe propose a stable clustering algorithm forVANETsWe propose a novel approach to form and maintainstable clusters for VANETs on highways to avoid continualcluster reforming A center-based clustering algorithm isused to locate the initial clustersrsquo centers In every clustera suitable CH is chosen by vehiclesrsquo position speed andmaximal acceleration A cluster maintenance algorithm isproposed to keep CMs in its CHrsquos transmission range

The rest of the paper is organized as follows The RelatedWork briefly reviews the current literature on clusteringalgorithms in VANETs The proposed scheme is detailed inthe Proposed SchemeThe simulation parameters simulationresults and analysis are shown in the Performance Evalua-tion In the Conclusion we state the conclusion

2 Related Work

In the literature clustering is the process to group vehicles inVANETS

Ref [5] proposes a method named LTE4V2X to orga-nize vehicular networks In the centralized vehicular net-works eNodeB manages vehicles in its coverage and dividesthem into clusters LTE4V2X protocol defines how the self-organized network works In LTE4V2X eNodeB creates clus-ters which contain the largest number of nodes circulating inthe same direction

Ref [6] extends LTE4V2X to increase information dis-semination efficiency It selects CH by the distance fromvehicles to eNodeB Although comparing to the originalapproach the complexity is lower and the LTE channel qual-ity is higher the power consumption of message exchangingis not optimized Nevertheless [6] states that the systemcan calculate the transmit power of DSRC channels by thedistance between vehicles so that the transmit power couldbe dynamically adjusted

Road condition affects the speed and direction of vehiclesFor example vehiclersquos speed is lower on the bumpy road thana smooth road Vehicle mobility is determined by humanbehavior Take a street connected megapolis and a village asan example In the morning most vehicles move from thevillage (home) to the megapolis (office) In the evening most

vehicles run following the reverse path Ref [7] quantifiestemporal locality similarity to measure the relation of twovehiclesrsquo mobility Then they utilize the relation of vehiclesrsquomovements to form stable clusters The locality can also beused for reducing energy consumption [8]

Ref [9] proposes a clustering approach to minimize thetotal power consumed by DSRC communications They usea weighted distance matrix to indicate power consumedbetween each pair of vehicles In this way the CH selectionproblem is formulated as a variant of the p-median problemin graph theory [10] In this approach the number of clusters119901 is determined first based on LTE coverage radius andDSRC coverage radiusThe119901 cluster zones are determined byvehicle number and 80211P coverage radius p Cluster Headsthat are closet to the eNodeB are selected Then the systemdynamically selects new CH to minimize the transmissionpower between CMs and CH based on weighted distance andthe p-median issue in graph theory Although this approachminimizes the power consumption within a single cluster thepower consumption of V2I communications has not beenconsidered The method to decide the zones is vague andcomplicated Moreover this approach is not suitable for thescenario that CMs not only send their information to CH butalso communicate among themselves

Ref [11] proposes a high-integrity file transfer schemefor VANETs on highways named Cluster-based File Transfer(CFT) scheme In this scheme CMs help their CH todownload file fragments and then transmit fragments tothe CH which requests the file Since the very high speedof vehicles on highways CFT is a good approach to helpthe vehicles download files which they have not enoughconnection time to download However CFT just considersthe bidirection environment In addition with CFT CHbroadcasts its request to its neighbors then neighbors thatreceive the invitation join the cluster and broadcast therequest to invite more vehicles to join the cluster untilthere are enough vehicles Therefore CFT may not able toapply in complicated environment and it may cause networkcongestions

Ref [12] proposes an evolutionary game theoretic (EGT)framework for clustering and CH selecting Their protocolis based on game theory They defined the net utility of aCH to select the CH which may achieve high throughput Acluster size is added in the utility function for CH to optimizethe size of a cluster Ref [13] proposes an intelligent naiveBayesian probabilistic estimation practice (ANTSC) methodThis method is based on the traffic flow To increase thestability of cluster a CH must be in the lane having theheaviest traffic flow Naive Bayes algorithm is used to selectthe CH which may make the cluster most stable However[12] just compared the EGT clustering with one clusteringalgorithm proposed in 2010 Both [12 13] did not considerthe security of the network

3 Proposed Scheme

31 Overview and Assumption Clustering algorithm groupsa set of unlabeled nodes into clusters In cluster-basedVANETs all vehicles send their information to eNodeB


eNodeB

CH

CM

LTE

802 11P












120579119896 = tanminus11199101015840119896minus 119910119896

1199091015840119896 minus 119909119896 (1)




















119863119896 =119873

sum119899=1



119881119896 =119873

sum119899=1

1003816100381610038161003816V119896 minus V1198991003816100381610038161003816 (3)


119860119896 =119873

sum119899=1

1003816100381610038161003816119886119896 minus 1198861198991003816100381610038161003816 (4)


M119896 = 120572 119863119896max 119863119899 | forall119899 isin 119862119894 + 120573 119881119896

max 119881119899 | forall119899 isin 119862119894+ 120574 119860119896

max 119860119899 | forall119899 isin 119862119894 (5)





1 SetM119898119894119899


119896

4 if119872119896 lt119872119898119894119899 then5 M

119898119894119899= M119896

6 119900 = 1198967 end8 end9 return o















119886value





119878119905119888 = 1119873 minus 1

119873minus1

sum119894=1

( Δ119905119894119897119894119891119890119905119894119898119890119894) (6)





119878119886119898 = sum119873minus1119894=1

119901119894 minus 119901119898sum119873minus1119894=1

119901119894 sdot 119873119873 minus 1 (7)















N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)






N=1

50

R=20

0N

=150

R=

300

N=1

50

R=40

0N

=300

R=

200

N=3

00

R=30

0N

=300

R=

400

N=4

50

R=20

0N

=450

R=

300

N=4

50

R=40

0

KMBSCalECBSC

0102030405060708090

100

Aver

age C

M L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0102030405060708090

Aver

age C

M L

ifetim

e (s)


0123456789

10

N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


0123456789

10

R=20

0v=

70

R=20

0v=

100

R=20

0v=

130

R=30

0v=

70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70

R=40

0v=

100

R=40

0v=

130

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003Pa

cket

Los

s Rat

e (

)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003

Pack

et L

oss R

ate (

)





5 Conclusion


Data Availability




References




























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



eNodeB

CH

CM

LTE

802 11P












120579119896 = tanminus11199101015840119896minus 119910119896

1199091015840119896 minus 119909119896 (1)




















119863119896 =119873

sum119899=1



119881119896 =119873

sum119899=1

1003816100381610038161003816V119896 minus V1198991003816100381610038161003816 (3)


119860119896 =119873

sum119899=1

1003816100381610038161003816119886119896 minus 1198861198991003816100381610038161003816 (4)


M119896 = 120572 119863119896max 119863119899 | forall119899 isin 119862119894 + 120573 119881119896

max 119881119899 | forall119899 isin 119862119894+ 120574 119860119896

max 119860119899 | forall119899 isin 119862119894 (5)





1 SetM119898119894119899


119896

4 if119872119896 lt119872119898119894119899 then5 M

119898119894119899= M119896

6 119900 = 1198967 end8 end9 return o















119886value





119878119905119888 = 1119873 minus 1

119873minus1

sum119894=1

( Δ119905119894119897119894119891119890119905119894119898119890119894) (6)





119878119886119898 = sum119873minus1119894=1

119901119894 minus 119901119898sum119873minus1119894=1

119901119894 sdot 119873119873 minus 1 (7)















N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)






N=1

50

R=20

0N

=150

R=

300

N=1

50

R=40

0N

=300

R=

200

N=3

00

R=30

0N

=300

R=

400

N=4

50

R=20

0N

=450

R=

300

N=4

50

R=40

0

KMBSCalECBSC

0102030405060708090

100

Aver

age C

M L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0102030405060708090

Aver

age C

M L

ifetim

e (s)


0123456789

10

N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


0123456789

10

R=20

0v=

70

R=20

0v=

100

R=20

0v=

130

R=30

0v=

70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70

R=40

0v=

100

R=40

0v=

130

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003Pa

cket

Los

s Rat

e (

)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003

Pack

et L

oss R

ate (

)





5 Conclusion


Data Availability




References




























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia














119863119896 =119873

sum119899=1



119881119896 =119873

sum119899=1

1003816100381610038161003816V119896 minus V1198991003816100381610038161003816 (3)


119860119896 =119873

sum119899=1

1003816100381610038161003816119886119896 minus 1198861198991003816100381610038161003816 (4)


M119896 = 120572 119863119896max 119863119899 | forall119899 isin 119862119894 + 120573 119881119896

max 119881119899 | forall119899 isin 119862119894+ 120574 119860119896

max 119860119899 | forall119899 isin 119862119894 (5)





1 SetM119898119894119899


119896

4 if119872119896 lt119872119898119894119899 then5 M

119898119894119899= M119896

6 119900 = 1198967 end8 end9 return o















119886value





119878119905119888 = 1119873 minus 1

119873minus1

sum119894=1

( Δ119905119894119897119894119891119890119905119894119898119890119894) (6)





119878119886119898 = sum119873minus1119894=1

119901119894 minus 119901119898sum119873minus1119894=1

119901119894 sdot 119873119873 minus 1 (7)















N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)






N=1

50

R=20

0N

=150

R=

300

N=1

50

R=40

0N

=300

R=

200

N=3

00

R=30

0N

=300

R=

400

N=4

50

R=20

0N

=450

R=

300

N=4

50

R=40

0

KMBSCalECBSC

0102030405060708090

100

Aver

age C

M L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0102030405060708090

Aver

age C

M L

ifetim

e (s)


0123456789

10

N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


0123456789

10

R=20

0v=

70

R=20

0v=

100

R=20

0v=

130

R=30

0v=

70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70

R=40

0v=

100

R=40

0v=

130

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003Pa

cket

Los

s Rat

e (

)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003

Pack

et L

oss R

ate (

)





5 Conclusion


Data Availability




References




























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia








119886value





119878119905119888 = 1119873 minus 1

119873minus1

sum119894=1

( Δ119905119894119897119894119891119890119905119894119898119890119894) (6)





119878119886119898 = sum119873minus1119894=1

119901119894 minus 119901119898sum119873minus1119894=1

119901119894 sdot 119873119873 minus 1 (7)















N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)






N=1

50

R=20

0N

=150

R=

300

N=1

50

R=40

0N

=300

R=

200

N=3

00

R=30

0N

=300

R=

400

N=4

50

R=20

0N

=450

R=

300

N=4

50

R=40

0

KMBSCalECBSC

0102030405060708090

100

Aver

age C

M L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0102030405060708090

Aver

age C

M L

ifetim

e (s)


0123456789

10

N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


0123456789

10

R=20

0v=

70

R=20

0v=

100

R=20

0v=

130

R=30

0v=

70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70

R=40

0v=

100

R=40

0v=

130

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003Pa

cket

Los

s Rat

e (

)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003

Pack

et L

oss R

ate (

)





5 Conclusion


Data Availability




References




























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia












N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

10

20

30

40

50

60

70

Aver

age C

H L

ifetim

e (s)






N=1

50

R=20

0N

=150

R=

300

N=1

50

R=40

0N

=300

R=

200

N=3

00

R=30

0N

=300

R=

400

N=4

50

R=20

0N

=450

R=

300

N=4

50

R=40

0

KMBSCalECBSC

0102030405060708090

100

Aver

age C

M L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0102030405060708090

Aver

age C

M L

ifetim

e (s)


0123456789

10

N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


0123456789

10

R=20

0v=

70

R=20

0v=

100

R=20

0v=

130

R=30

0v=

70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70

R=40

0v=

100

R=40

0v=

130

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003Pa

cket

Los

s Rat

e (

)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003

Pack

et L

oss R

ate (

)





5 Conclusion


Data Availability




References




























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



N=1

50

R=20

0N

=150

R=

300

N=1

50

R=40

0N

=300

R=

200

N=3

00

R=30

0N

=300

R=

400

N=4

50

R=20

0N

=450

R=

300

N=4

50

R=40

0

KMBSCalECBSC

0102030405060708090

100

Aver

age C

M L

ifetim

e (s)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0102030405060708090

Aver

age C

M L

ifetim

e (s)


0123456789

10

N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


0123456789

10

R=20

0v=

70

R=20

0v=

100

R=20

0v=

130

R=30

0v=

70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70

R=40

0v=

100

R=40

0v=

130

Aver

age N

umbe

r of R

eaffi

liatio

nTi

mes

per

Veh

icle

KMBSCalECBSC


N=1

50

R=20

0

N=1

50

R=30

0

N=1

50

R=40

0

N=3

00

R=20

0

N=3

00

R=30

0

N=3

00

R=40

0

N=4

50

R=20

0

N=4

50

R=30

0

N=4

50

R=40

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003Pa

cket

Los

s Rat

e (

)


R=20

0v=

70R=

200

v=10

0R=

200

v=13

0R=

300

v=70

R=30

0v=

100

R=30

0v=

130

R=40

0v=

70R=

400

v=10

0R=

400

v=13

0

KMBSCalECBSC

0

0005

001

0015

002

0025

003

Pack

et L

oss R

ate (

)





5 Conclusion


Data Availability




References




























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia





5 Conclusion


Data Availability




References




























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia














RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


a center-based secure and stable clustering algorithm for...

Documents