Research Article Adaptive Collaborative Detection for ...downloads.hindawi.com/journals/ijdsn/2014/275609.pdfdetection focuses on the retrieve of useful data among vehicles. In this

Research ArticleAdaptive Collaborative Detection for Opportunistic VehicleSensor Networks

Yuanyuan Zeng12

1 School of Electronic Information Wuhan University Wuhan 430072 China2 Su Zhou Institute of Wuhan University Suzhou 215123 China

Correspondence should be addressed to Yuanyuan Zeng zengyywhueducn

Received 31 October 2013 Revised 12 April 2014 Accepted 15 April 2014 Published 5 May 2014

Academic Editor Mohamed Tawhid

Copyright copy 2014 Yuanyuan Zeng This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Transportation is a huge problem that curbs development of societies and economy in many countries nowadays Participatorysensing technologies encourage people to be involved in environment monitoring through smart devices Vehicle sensor networks(VSNs) are a novel solution for road event detection with advantages In this paper we consider runtime road event detectionusing VSNs to satisfy the application-specific requirements through collaboration among vehicles A group road detection scheme(GRD) by using dynamic clustering in VSNs is presented to improve detection performance with low time complexity andmessagecomplexity The simulations with testbed of 5 remote controllable vehicles show that GRD scheme provides effectiveness underdifferent road scenarios

1 Introduction

Transportation problems become more and more imminentinmany countries nowadays Road detection is necessary andimportant for road safety and good transportation Participa-tory sensing encourages people to take part in environmentmonitoring by smart devices that they carry It helps to realizepervasive sensing and computing through public monitoringand information sharing

A vehicle sensor network is a kind of special mobilewireless sensor networks for vehicle environment [1] Sensordevices equipped on vehicles constitute a vehicle basedmobile sensor network The VSN senses data opportunis-tically and collects data through ad hoc communicationsamong vehicles Consider the mobility and coverage ofvehicles among a city without extra overhead a VSN providesa novel and convenient solution for road event detectionConsidering high mobility of vehicles however collectedsensing data in a VSN is often too sparse to use To improvethe detection performance group detection is a feasiblesolution based on collaboration among vehicles

In this paper we consider runtime road detection inVSNsto improve application-specific detection performance Wepresent a group road detection scheme (GRD) by utilizing

dynamic collaborative clustering In GRD the nodes dynam-ically join and leave the cluster according to the detectionrequirements Each node in a cluster collaborates togetherto detect a road event The detection is required to achieveapplication-specific detection performance without muchoverheadThere is a tradeoff between detection capability andcomplexity

The main contribution of this paper includes the fol-lowing (1) The novel group detection scheme in VSNs isproposed to improve detection performance (2) Dynamicalclustering algorithms are proposed to form groups andachieve group detection (3) The proposed algorithms aredistributed with low message and time complexity

Section 2 presents the relatedwork Section 3 presents thenetwork model We present the details of GRD in Section 4Section 5 is simulations and evaluations Section 6 concludesthis paper

2 Related Works

Participatory sensing is a revolutionary research directionwhich allows people with smart device to participate inenvironment detection and monitoring voluntarily In thissituation the sensed data will be processed and shared

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2014 Article ID 275609 8 pageshttpdxdoiorg1011552014275609

2 International Journal of Distributed Sensor Networks

Vehicles with sensors

Internet

Environment

Temperature

Humidity

Tail gas

Vehicles

Time

Location

Heading

Speed

Accelerators

Events

Wetslippery

Bumping

Traffic jam

Road side unit

Vehicle sensornetworksSurveillance

center

middot middot middot



Figure 1 A vehicle sensor network for road event detection applications

through wireless networks and the other basic infrastruc-tures Vehicle sensor networks [2] are the combined outcomeof vehicle ad hoc networks and WSNs VSNs could be usedin a wide range of applications [3] for example in trafficmonitoring to save communication costs and avoid networkcongestions Vehicle based sensors could also be used tomonitor and identify possible vehicle thefts by detectingunauthorized vehicle movements

In recent research work of sensor networks road surfacemonitoring related applications are popular Mohan et al[4] propose system solution to monitoring road and trafficconditions in a city which aims to detect potholes bumpsbraking and honking by using smart phones with theaccelerometer microphone GSM radio and GPS sensorsKaruppuswamy et al [5] propose a vision based scheme forpothole avoidance in a mobile robot by the use of camerasThey detect potholes by looking for large circular objects inthe field of view Eriksson et al [6] propose road surfacepothole detection by using machine learning mechanismbased on hidden Markov model Keally et al [7] propose aconfident event detection algorithm based on hiddenMarkovmodel in terms of application-specific detection demand

For vehicular networks clustering is a necessary methodto cluster and manage vehicles [8] Rawashdeh and Mahmud[9] propose a speed-overlapped clustering technique onhighways by using vehicular speed and relative movementdirection Gurung et al [10] propose a moving-zone basedarchitecture with moving object modeling and indexingtechniques to facilitatemessage routing inVANETs accordingto the unique characteristics of network environments Ucaret al [11] propose a clustering mechanism to choose thenode with the least mobility calculated by speed differenceamong neighbor nodes as the cluster head to maintain thestable clustering in VANETs Most of the above work is forvehicular ad hoc networks and the clustering scheme aims

to communications among vehicles None of them is foreffective road event detection Communications focus onthe effective connections and information exchange whiledetection focuses on the retrieve of useful data amongvehicles

In this context the paper focuses on dynamically-clustered collaboration based group road detection schemein vehicle sensor networks with low overhead

3 System Models

Vehicle sensor networks are widely used in the area of intelli-gent transportation systems related applications and here thepaper focuses on road event detection applications Vehiclescollect road event data by opportunistic communications asshown in Figure 1

We use a dynamic graph 119866(119881 119864 119879) to model the VSNsEach vehicle V isin 119881 is a graph node Here each vehicle (graphnode) can include more than one sensor to detect the roadevents The graph edge 119890 isin 119864 represents the opportunisticwireless communication link between vehicles Because ofthe time-dependent opportunistic communication the con-necting time 119905 is used to record the vehicle communicationtrajectory

In road detection process each vehicle can be in one ofthe states initialized clustering and clustered as in Figure 2

ldquoInitializedrdquo is the initial state for vehicle when it is not inany of the groups An invoker broadcasts out a cluster-requestmessage to begin the dynamic clustering process

ldquoClusteringrdquo is the state when a node is in the decisionof whether to be in the current cluster or not Either aninvoker or the other vehicles can be in this state If an invokerreceives responses from the other nodes that are willing tocooperate the clustering process is successful and then theinvoker enters into ldquoclusteredrdquo Otherwise the vehicle will

International Journal of Distributed Sensor Networks 3

Initialized

Clustering

Clustered

Invoker or receivecluster request message Clustering

fails

Clusteringsucceeds

Cluster baseddetection is over or

timeout

Figure 2 State transition machine for a vehicle

enter back to ldquoinitializedrdquo If it is a noninvoker vehicle that isin ldquoclusteringrdquo it will calculate to make decision on whetherto join or to reject the clustering request A noninvokernode which decides to join the current cluster will enter intoldquoclusteredrdquo Otherwise it enters into ldquoinitializedrdquo

Obviously ldquoclusteredrdquo is the state that a node is in thecluster and cooperates with the other clustermembersWhenthe cluster based group detection is over or larger than apredetermined timeout the node in ldquoclusteredrdquo enters backto ldquoinitializedrdquo state and leaves the cooperation The timeoutis used to be released from long time utilization or lostconnection from the clustering If a vehicle joins the groupdetection it will keep staying in the group and work as agroup member for a while The maximal working time istimeout

4 GRD Design

GRD design aims to construct dynamic groups (clusters) toguarantee efficient road detection The dynamic cluster isconstructed through the following function modules localinvoker cluster generation and runtime decision modules

ldquoLocal invokerrdquo is the module to determine the localinvoker for GRD The local invoker is the vehicle with sparsedata that needs to utilize GRD to improve the detection Theinvoker sends out a group detection request (ie cluster-request message) to request for clustering

ldquoCluster generationrdquo is the module to choose event-related vehicles to join into groups and try to satisfy thedetection requirement

ldquoRuntime decisionrdquo module makes decisions for thevehicle on whether to joinform the group or not

Figure 3 illustrates the above modules on both invokerand noninvoker vehicles respectively The noninvoker vehi-cles join the cluster or reject the request through themodulesIt can also terminate current cluster based group detectionaccording to application requirements or local detectionefficiency

41 The Local Invoker For the convenience of illustrationsome of the notations used in the paper are explained inNotation Section

Let the road detection zone 119885 which can be modeled byan approximate rectangle area and divided into subzone asgrids 119892

1 1198922 119892

119898in the whole surveillance area that is119885 =

1198921 1198922 119892

119898 For detection grid 119892

119894 the centroid location is

denoted by 119911119894 Let 119889

119896119894be the distance between a vehicle 119899

119896

and event related grid 119892119894 In (1) 119899

119896(119909 119910) is the coordinates of

vehicle 119899119896and 119911119894(119909 119910) is the coordinates of grid centroid

119889119896119894equiv 119889 (119899

119896 119892119894) =

1003817100381710038171003817119899119896 (119909 119910) minus 119911119894 (119909 119910)1003817100381710038171003817 (1)

119888(119899119896 119892119894) denotes the static detection capability of vehicle 119899

119896

on a specific static position toward the road grid 119892119894 The

vehicle can be equipped with sensors to detect road eventsThe maximal sensing range of the vehicle 119899

119896is denoted as

119877119896 A probabilistic model is utilized to calculate the detection

capability of the vehicle toward a specific event grid byconsidering the distance between the vehicle and the gridcentroid It is obvious that a vehicle has better detectioncapability when it is near the grid 119888(119899

119896 119892119894) can be illustrated

as follows

119888 (119899119896 119892119894) =

1

1 + radic119889119896119894

119889119896119894le 119877119896

0 119889119896119894gt 119877119896

(2)

The vehicle detection capability can be calculated accordingto (2) as a value between 0 and 1 So it is convenient for upper-layer applications to give a predefined capability parameteror threshold according to application requirements and long-time observation data

Considering the vehicle mobility the mobile detectioncapability (to abbreviate it as detection capability in theremaining) for vehicles should include the summation of allthe static detection capability over time and space 1198881015840(119899

119896 119892119894)

is used to denote the detection capability of vehicle 119899119896toward

the road event grid 119892119894 when it moves in grid 119892

119894 In (3)

(119909min 119910min) and (119909max 119910max) are the coordinate range ofthe vehicle movement and (119905

1 1199052) is the time durations of

movement

1198881015840(119899119896 119892119894) = int1199052

1199051

int119909max

119909min

int119910max

119910min

119888 (119899119896 119892119894) 119889119905 119889119909 119889119910 (3)

If a vehicle moves across the grids the detection capabilityfor a specific road event is calculated only through theevent-related grids during the movement Figure 4 showsan example In the example the vehicle moves across theupper-right side two grids 119892

4and 119892

5 But if only 119892

5is event-

related grid the detection capability of the vehicle will becalculated according to themovement within the grid 119892

5The

event-related grid is decided by the event detection precisionthat is usually defined by the upper-layer application with adetection precision range (which can be modeled as a circlewith a predefined radius)The grids that fall into the precisionrange are the event-related grids

If a vehicle finds the local detection capability not suitablefor the application-specific detection performance it willinvoke the group detection process and becomes an invoker

According to the state transition machine the localinvoker broadcasts out a cluster request with the request ofclustering process and enters into ldquoclusteringrdquo The cluster-request message broadcast is based on CSMACA mecha-nism to avoid contentions The node in ldquoinitializedrdquo state


Localinvoker

Invoker Noninvoker

Clustergeneration

Runtimedecision

Clustergeneration

Runtimedecision

Joinrejectterminate

Groupdetectionrequest

Figure 3 Dynamic cluster related function modules

Moving trajectory Event Vehicle

Moving area in event grid

Zoom in

g1 g2 g3 g4

g4

g5 g6 g7 g8

g8

g9 g10 g11 g12

Figure 4 An example of detection scenario when vehicles moveacross the grids

that receives a cluster request message enters into ldquoclusteringrdquostate which is the state to make decision on cooperation

Algorithm 1 describes Pseudocode of the invoker selec-tion The message is broadcasted among the neighborhoodEach node executes the algorithm in sequence in a distribu-tive way It is obvious that Algorithm 1 is with119874(|Δ|)messagecomplexity and119874(119899) time complexity whereΔ is themaximaldegree of the network graph

42 Cluster Generation After the invoker broadcasts out acluster request the neighbors in ldquoinitializedrdquo state whichreceives themessage will decide whether to join the requestedgroup detection or not by runtime decision module If nodesin other states receive the message they just ignore it Thecluster requestmessagewill be forwarded in the network untilthe group achieves satisfied detection capability

Define the group detection capability of vehicle group set119878 invoked by 119899

119896for a specific road event which occurs in grid

119892119894as 1198881015840(119878

119896 119892119894) In (4) 1 minus 1198881015840(119899

119895 119892119894) is the detection disability

degree of vehicle 119899119895toward grid 119892

119894 Then the disability of the

set 119878 is achieved by the multiplier of each vehiclersquos disabilityBased on it the group detection capability is shown as in

1198881015840(119878119896 119892119894) = 1 minus Π

119899119895isin119878119896

(1 minus 1198881015840(119899119895 119892119894)) (4)

Algorithm 2 describes cluster generation process To curb thenumber and scale of forwardingmessages amaximal forward

number MAX FW is defined to avoid excessive broadcastmessages in the network It is easy to see that Algorithm 2is the same as Algorithm 1 with low timemessage complexityand suitable for VSNs

43 Dynamically Clustered Cooperation and Decision Inruntime decision process for an invoker vehicles that cancontribute to current event detection are invited to joinin to improve detection capabilities For a noninvoker itwill make decision according to the detection relevancewith the inviter So it needs a negotiation process betweenruntime dynamic-clustered cooperation between the invokerand the noninvoker Through the negations among vehiclesnonrelevant vehicles are filtered from the cluster and relevantvehicles are involved in the cluster to make group detectionwith the application-specific detection capability

The detection relevance metric is defined to determinewhat kind of degree the vehicle can contribute to the currentevent detection Assume invoker 119899

119896and ldquoclusteringrdquo vehicle

119899119895get observed dataduring time duration 119879 from start time

1199051to end time 119905

2The detection relevance of 119899

119895with cluster119862

119896

toward event grid 119892119894is denoted as 119903

119894(119862119896 119899119895) which is defined

as shown in (5) in which 1198881015840(119862119896 119892119894) is the detection capability

of cluster119862119896toward event grid 119892

119894during time durationT and

1198881015840(119899119895 119892119894) is the detection capability of vehicle 119899

119895toward 119892

119894

during T

119903119894(119862119896 119899119895) =

1198881015840 (119862119896cup 119899119895 119892119894)

1198881015840 (119862119896 119892119894)

when 1198881015840 (119862119896 119892119894) gt 0 (5)

Especially before the cluster is constructed the detectionrelevance of 119899

119895with 119899

119896toward event grid 119892

119894is denoted as

119903119894(119899119896 119899119895) as in

119903119894(119899119896 119899119895) =

1198881015840 (119899119896 cup 119899

119895 119892119894)

1198881015840 (119899119896 119892119894)

when 1198881015840 (119899119896 119892119894) gt 0

(6)

If 119903119894(119862119896 119899119895) gt 1 or 119903

119894(119899119896 119899119895) gt 1 it implies that 119899

119895

will contribute to the event detection for current cluster orpossible cluster In this way the detection relevance can becalculated The relevant vehicles join into the cluster or leavethe cluster to achieve efficient detection Algorithm 3 is thepseudocode of runtime decision It is obvious to see thatthe algorithm is with linear message complexity and timecomplexity


Input a VSN 119866(119881 119864 119879) application-specific detection capability 119888th initial state ldquoinitializedrdquoOutput nodes for clustering processfor each node in 119866

If (1198881015840(119892119894 119899119896) gt 0)(1198881015840(119892

119894 119899119896) lt 119888th) ampamp (state = ldquoinitializedrdquo)

Ignore received messagesbroadcast a Cluster-Req messagestatelarr ldquoclusteringrdquo

elseif receives a Cluster-Req ampamp (state = ldquoinitializedrdquo)statelarr ldquoclusteringrdquo

endifendfor

Algorithm 1 Invoker algorithm

Input a VSN 119866(119881 119864 119879) application-specific detection capability 119888th nodes for clustering by Algorithm 1Output Set of clusters 119862for each node in 119866If (state = ldquoclusteringrdquo) ampamp (non-invoker) ampamp (receive ldquojoinrdquo msgs) Runtime Decision() is the Runtime Decision algorithm

statelarr ldquoclusteredrdquobecome a cluster member

elseif (Runtime Decision()) reject the clustering or terminate from the cluster

statelarr ldquoinitializedrdquoIf (msg seq lt= MAX FW)

forward the Cluster Req msgendif

endifIf (state = ldquoclusteringrdquo) ampamp (invoker)

If (wait(T MAX1)) ampamp (receive ldquojoinrdquo responses) wait(T MAX1) is the delay function for clustering T MAX1 is the maximal waiting time for clustering

statelarr ldquoclusteredrdquobecome a clusterIf 1198881015840(119862

119896 119892119894) lt 119888th

119862119896is the group set of current cluster invoked by 119899

119896

Broadcast a Cluster Req msgendif

elseif (wait(T MAX1)) (receive ldquorejectrdquo)statelarr ldquoinitializedrdquo

no ldquojoinrdquo received during waiting time implies failureendif

endifIf (state = ldquoclusteredrdquo) ampamp (invoker) ampamp (receive ldquojoinrdquoldquoterminaterdquo msgs)

Update the clusterendif

endfor

Algorithm 2 Cluster generation

5 Simulations

The overall goal of this part is to make simulations in theremote-control vehicle based transportation situations andthen make evaluations to testify the effectiveness and accu-racy of GRD Actually the detection performance is relatedto the detection methods The simulations are implementedbased on the same basic detection method with the situation

of only local vehicle detection involved and with GRDrespectively The comparisons are made that try to show thedifference between the two schemes based on the same basicdetectionmethod To simplify the tests and show the obviousdifferences from the results a threshold based event detectionmethod is taken in the simulations In the simulationslocal vehicle detection scheme is to make basic thresholdbased event detectionwithout the group collaboration among


Input a VSN 119866(119881 119864 119879) application-specific detection capability 119888th Set of Growing clusters 119862 (the Step output of Algorithm 2)Output node responses for clustersfor each node in 119866

If (state = ldquoclusteringrdquo) ampamp ((119903119894(119862119896 119899119895) gt 1) (119903

119894(119899119896 119899119895) gt 1))

Response to invoker with msg(join)elseResponse to invoker with msg(reject)

endifIf (state = ldquoclusteredrdquo) ampamp application needs

cluster members can leave the group for application needsResponse to invoker with msg(terminate)

endifendfor

Algorithm 3 Runtime decision algorithm

Figure 5 A remote controllable vehicle

vehicles that is the sensing data of each vehicle is comparedwith the application-specific threshold and then to decide theevent based on it The GRD scheme proposed in the paperis to make threshold based event detection with the groupcollaboration scheme that is the average sensing data ofthe group achieved is compared with the application-specificthreshold and then to decide the event

A simulated testbed consists of 5 remote controllablevehicles The vehicle can work over 30 minutes continuouslyThe vehicle for testbed is shown in Figure 5

An illustration of testbed is shown in Figure 6The vehicleis based on Arduino control board with 4 wheel drive It isequipped with accelerometers during the simulations Thevehicle couldmove flexibly as in real transportationThe sen-sors on the vehicles are connected to one another wirelesslythrough the low-power radio Two vehicles in the testbedare also equipped with a WiFi card The test area is coveredby WiFi Then the sensing data collected by each vehicleis forwarded and then gathered to central gateway throughWiFi connectionThedriving area is divided into virtual gridsand vehicles drive according to the grid lane Each vehiclecan drive toward the vertical and horizontal directions asshown in the figure The vehicle can be controlled to change

VehicleGrid lane

Direction vector

Figure 6 An Illustration of driving scenario in testbed

the direction when it approaches the area border or to avoidcollisions

Each vehicle is placed randomly on different loca-tions and lanes and then controlled to drive in differentdirections around the test area There are several artificialupwarddownward slopes built by boards and slippery roadareamade by spilling jelly as well as uneven areas with brokenstones in the test place to simulate complex road drivingTheartificial experiment scenarios are used to simulate unevenwet and slippery road conditions The location of simulatedroad events is randomly chosen Vehicles are controlled tomove around in the test place The sensors are mountedon each vehicle with the same orientation and place Theultrasonic wave sensor is used for measuring the distanceand location In the simulations the vehicles are controlledto travel within 5 different road condition scenarios as shownin Table 1

Figure 7 shows the detection accuracy rate within differ-ent scenariosThe results showwe achieve 253ndash402moreaccuracy ratio when using group detection In GRD morerelevant vehicles are involved into detection and improve


Table 1 Simulated road scenarios

Scenarios Number of uneven areas Number of slippery areas Number of slopesScenario 1 5 2 1Scenario 2 2 5 1Scenario 3 8 1 2Scenario 4 1 8 2Scenario 5 5 5 3

1 2 3 4 5606264666870727476788082848688909294

Local vehicle detectionGRD

Accu

racy

ratio

()

Road condition scenario number

Figure 7 Detection accuracy ratio within different scenarios

1 2 3 4 5

14

16

18

20

22

24

26

28

30

False

pos

itive

ratio

()



Figure 8 False positive ratio within different scenarios

the accuracy by more useful observation data GRD helps toimprove the accuracy of detection

Figure 8 shows false positive rate within different scenar-ios GRD helps to improve the false positive rate

Figure 9 shows the false negative rate within different sce-narios GRD decreases around 52 false negative Obviously

1 2 3 4 50

5

10

15

20

25

30

False

neg

ativ

e rat

io (

)



Figure 9 False negative rate within different scenarios

1 2 3 4 5

10

15

20

25

30

35

40

45

Ave r

esid

ual e

nerg

y ra

tio (

)



Figure 10 Average residual energy ratio within different scenarios

group detection has better performance than only using localvehicle detection

Figure 10 shows the average residual energy ratio ofnetwork nodeswithin different scenariosThe residual energyratio is calculated by residual energy versus initial energyWecompare GRDwith local vehicle detection GRD results in no


more than 17 average residual energy ratio when comparedwith local vehicle detection cases in the network SoGRDwillnot incur too much energy consumption Especially for realvehicles the sustainable energy supply is not a big problem

6 Conclusion

Vehicle sensor networks will be widely utilized in intelligenttransportation systems in a very near future which can beused to inform the drivers about kinds of road conditionevents In this paper we propose a group detection schemebased on dynamically clustered cooperation scheme calledGRD in VSNs GRD scheme includes algorithms of clusterinvoked process cluster generation and runtime decisionthrough vehicle collaborations The simulations are carriedout on 5 controllable vehicles with different scenarios Theresults show that GRD can help to improve detection perfor-mance effectively without much overhead

The further work will also be carried out on the testbedconsisting of voluntary cars equippedwith sensors on campusroad areas The test results will also be compared with otherrelated work to testify the effectiveness

Notation

119885 The approximated rectangle roaddetection zone

119892119894 The divided subgrid in 119885

119911119894 The centroid location of 119892

119894

119899119896 The vehicle with its ID as 119896

119877119896 The sensing range of 119899

119896

119889119896119894 The distance between 119899

119896and 119892

119894

119888(119899119896 119892119894) The static detection capability of 119899

119896toward

119892119894

1198881015840(119899119896 119892119894) The mobile detection capability of 119899

119896

toward 119892119894

119878 The vehicle group set119862119896 The cluster invoked by 119899

119896

1198881015840(119878119896 119892119894) The detection capability of 119878 invoked by 119899

119896

toward 119892119894

119903119894(119862119896 119899119895) The detection relevance of 119899

119895with 119862

119896

toward 119892119894

Conflict of Interests

The author declares that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

Thanks are due to the anonymous reviewers and the academiceditorThis work was supported by Chinese National ScienceFoundation Project (no 61103218)TheNatural Science Foun-dation of Jiangsu Province Youth Project (no BK2012200)Hubei Province National Science Foundation Project (no2011CDB446) and Scientific Research Foundation for theReturned Overseas Chinese Scholars State Education Min-istry

References

[1] M Gerla and L Kleinrock ldquoVehicular networks and the futureof the mobile internetrdquo Computer Networks vol 55 no 2 pp457ndash469 2011

[2] K Lan C Chou and H Wang ldquoUsing vehicular sensornetworks for mobile surveillancerdquo in Proceedings of the 76thIEEE Vehicular Technology Conference (VTC Fall) pp 1ndash5Quebec City Canada September 2012

[3] H Hartenstein and K Laberteaux VANET Vehicular Applica-tions and Inter-Networking Technologies John Wiley amp Sons2010

[4] P Mohan V N Padmanabhan and R Ramjee ldquoTraffic-sense rich monitoring of road and traffic conditions usingmobile smartphonesrdquo Tech Rep MSR-TR-2008-59 MicrosoftResearch 2008

[5] J Karuppuswamy V Selvaraj M M Ganesh and E L HallldquoDetection and avoidance of simulated potholes in autonomousvehicle navigation in an unstructured environmentrdquo in Intelli-gent Robots and Computer Vision XXI Algorithms Techniquesand Active Vision vol 4197 of Proceedings of SPIE pp 70ndash80November 2000

[6] J Eriksson L Girod B Hull R Newton S Madden andH Balakrishnan ldquoThe Pothole Patrol using a mobile sensornetwork for road surface monitoringrdquo in Proceedings of the 6thInternational Conference on Mobile Systems Applications andServices pp 29ndash39 June 2008

[7] M Keally G Zhou and G Xing ldquoWatchdog confident eventdetection in heterogeneous sensor networksrdquo in Proceedingsof the 16th IEEE Real-Time and Embedded Technology andApplications Symposium (RTAS rsquo10) pp 279ndash288 April 2010

[8] S Vodopivec J Bester and A Kos ldquoA survey on clusteringalgorithms for vehicular Ad-Hoc networksrdquo in Proceedings ofthe 35th International Conference on Telecommunications andSignal Processing (TSP rsquo12) pp 52ndash56 July 2012

[9] Z Rawashdeh and S Mahmud ldquoA novel algorithm to formstable clusters in vehicular ad hoc networks on highwaysrdquoEURASIP Journal onWireless Communications and Networkingvol 2012 article 15 pp 1ndash13 2012

[10] S Gurung D Lin A Squicciarini andO K Tonguz ldquoAmovingzone based architecture formessage dissemination inVANETsrdquoin Proceedings of the 8th International Conference on Networkand Service Management Workshop on System VirtualizationManagement pp 184ndash188 October 2012

[11] S Ucar S C Ergen and O Ozkasap ldquoVMaSC vehicularmulti-hop algorithm for stable clustering in vehicular Ad Hocnetworksrdquo in Proceedings of the IEEE Wireless Communicationsand Networking Conference (WCNC rsquo13) pp 2381ndash2386 April2013

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

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



Vehicles with sensors

Internet

Environment

Temperature

Humidity

Tail gas

Vehicles

Time

Location

Heading

Speed

Accelerators

Events

Wetslippery

Bumping

Traffic jam

Road side unit

Vehicle sensornetworksSurveillance

center




Figure 1 A vehicle sensor network for road event detection applications

through wireless networks and the other basic infrastruc-tures Vehicle sensor networks [2] are the combined outcomeof vehicle ad hoc networks and WSNs VSNs could be usedin a wide range of applications [3] for example in trafficmonitoring to save communication costs and avoid networkcongestions Vehicle based sensors could also be used tomonitor and identify possible vehicle thefts by detectingunauthorized vehicle movements

In recent research work of sensor networks road surfacemonitoring related applications are popular Mohan et al[4] propose system solution to monitoring road and trafficconditions in a city which aims to detect potholes bumpsbraking and honking by using smart phones with theaccelerometer microphone GSM radio and GPS sensorsKaruppuswamy et al [5] propose a vision based scheme forpothole avoidance in a mobile robot by the use of camerasThey detect potholes by looking for large circular objects inthe field of view Eriksson et al [6] propose road surfacepothole detection by using machine learning mechanismbased on hidden Markov model Keally et al [7] propose aconfident event detection algorithm based on hiddenMarkovmodel in terms of application-specific detection demand

For vehicular networks clustering is a necessary methodto cluster and manage vehicles [8] Rawashdeh and Mahmud[9] propose a speed-overlapped clustering technique onhighways by using vehicular speed and relative movementdirection Gurung et al [10] propose a moving-zone basedarchitecture with moving object modeling and indexingtechniques to facilitatemessage routing inVANETs accordingto the unique characteristics of network environments Ucaret al [11] propose a clustering mechanism to choose thenode with the least mobility calculated by speed differenceamong neighbor nodes as the cluster head to maintain thestable clustering in VANETs Most of the above work is forvehicular ad hoc networks and the clustering scheme aims

to communications among vehicles None of them is foreffective road event detection Communications focus onthe effective connections and information exchange whiledetection focuses on the retrieve of useful data amongvehicles

In this context the paper focuses on dynamically-clustered collaboration based group road detection schemein vehicle sensor networks with low overhead

3 System Models

Vehicle sensor networks are widely used in the area of intelli-gent transportation systems related applications and here thepaper focuses on road event detection applications Vehiclescollect road event data by opportunistic communications asshown in Figure 1

We use a dynamic graph 119866(119881 119864 119879) to model the VSNsEach vehicle V isin 119881 is a graph node Here each vehicle (graphnode) can include more than one sensor to detect the roadevents The graph edge 119890 isin 119864 represents the opportunisticwireless communication link between vehicles Because ofthe time-dependent opportunistic communication the con-necting time 119905 is used to record the vehicle communicationtrajectory

In road detection process each vehicle can be in one ofthe states initialized clustering and clustered as in Figure 2

ldquoInitializedrdquo is the initial state for vehicle when it is not inany of the groups An invoker broadcasts out a cluster-requestmessage to begin the dynamic clustering process

ldquoClusteringrdquo is the state when a node is in the decisionof whether to be in the current cluster or not Either aninvoker or the other vehicles can be in this state If an invokerreceives responses from the other nodes that are willing tocooperate the clustering process is successful and then theinvoker enters into ldquoclusteredrdquo Otherwise the vehicle will


Initialized

Clustering

Clustered


fails

Clusteringsucceeds


timeout




4 GRD Design








1 1198922 119892


1198921 1198922 119892



denoted by 119911119894 Let 119889


119896




119889119896119894equiv 119889 (119899

119896 119892119894) =

1003817100381710038171003817119899119896 (119909 119910) minus 119911119894 (119909 119910)1003817100381710038171003817 (1)


119896



119896is denoted as




as follows

119888 (119899119896 119892119894) =

1

1 + radic119889119896119894

119889119896119894le 119877119896

0 119889119896119894gt 119877119896

(2)



119896 119892119894)



119894 In (3)



movement

1198881015840(119899119896 119892119894) = int1199052

1199051

int119909max

119909min

int119910max

119910min

119888 (119899119896 119892119894) 119889119905 119889119909 119889119910 (3)


4and 119892


5is event-


5The





Localinvoker

Invoker Noninvoker

Clustergeneration

Runtimedecision

Clustergeneration

Runtimedecision

Joinrejectterminate





Zoom in

g1 g2 g3 g4

g4

g5 g6 g7 g8

g8

g9 g10 g11 g12







119892119894as 1198881015840(119878

119896 119892119894) In (4) 1 minus 1198881015840(119899





1198881015840(119878119896 119892119894) = 1 minus Π

119899119895isin119878119896

(1 minus 1198881015840(119899119895 119892119894)) (4)







1199051to end time 119905



119896


119894(119862119896 119899119895) which is defined





119895toward 119892

119894

during T

119903119894(119862119896 119899119895) =

1198881015840 (119862119896cup 119899119895 119892119894)

1198881015840 (119862119896 119892119894)

when 1198881015840 (119862119896 119892119894) gt 0 (5)


119895with 119899


119894is denoted as

119903119894(119899119896 119899119895) as in

119903119894(119899119896 119899119895) =

1198881015840 (119899119896 cup 119899

119895 119892119894)

1198881015840 (119899119896 119892119894)

when 1198881015840 (119899119896 119892119894) gt 0

(6)

If 119903119894(119862119896 119899119895) gt 1 or 119903


119895




If (1198881015840(119892119894 119899119896) gt 0)(1198881015840(119892




endifendfor










119896 119892119894) lt 119888th


119896






endfor


5 Simulations






119894(119899119896 119899119895) gt 1))




endifendfor






VehicleGrid lane

Direction vector








1 2 3 4 5606264666870727476788082848688909294


Accu

racy

ratio

()



1 2 3 4 5

14

16

18

20

22

24

26

28

30

False

pos

itive

ratio

()







1 2 3 4 50

5

10

15

20

25

30

False

neg

ativ

e rat

io (

)




1 2 3 4 5

10

15

20

25

30

35

40

45

Ave r

esid

ual e

nerg

y ra

tio (

)








6 Conclusion



Notation




119894



119896


119896and 119892

119894


119896toward

119892119894


119896

toward 119892119894


119896


119896

toward 119892119894


119895with 119862

119896

toward 119892119894



Acknowledgments


References














RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Initialized

Clustering

Clustered


fails

Clusteringsucceeds


timeout




4 GRD Design








1 1198922 119892


1198921 1198922 119892



denoted by 119911119894 Let 119889


119896




119889119896119894equiv 119889 (119899

119896 119892119894) =

1003817100381710038171003817119899119896 (119909 119910) minus 119911119894 (119909 119910)1003817100381710038171003817 (1)


119896



119896is denoted as




as follows

119888 (119899119896 119892119894) =

1

1 + radic119889119896119894

119889119896119894le 119877119896

0 119889119896119894gt 119877119896

(2)



119896 119892119894)



119894 In (3)



movement

1198881015840(119899119896 119892119894) = int1199052

1199051

int119909max

119909min

int119910max

119910min

119888 (119899119896 119892119894) 119889119905 119889119909 119889119910 (3)


4and 119892


5is event-


5The





Localinvoker

Invoker Noninvoker

Clustergeneration

Runtimedecision

Clustergeneration

Runtimedecision

Joinrejectterminate





Zoom in

g1 g2 g3 g4

g4

g5 g6 g7 g8

g8

g9 g10 g11 g12







119892119894as 1198881015840(119878

119896 119892119894) In (4) 1 minus 1198881015840(119899





1198881015840(119878119896 119892119894) = 1 minus Π

119899119895isin119878119896

(1 minus 1198881015840(119899119895 119892119894)) (4)







1199051to end time 119905



119896


119894(119862119896 119899119895) which is defined





119895toward 119892

119894

during T

119903119894(119862119896 119899119895) =

1198881015840 (119862119896cup 119899119895 119892119894)

1198881015840 (119862119896 119892119894)

when 1198881015840 (119862119896 119892119894) gt 0 (5)


119895with 119899


119894is denoted as

119903119894(119899119896 119899119895) as in

119903119894(119899119896 119899119895) =

1198881015840 (119899119896 cup 119899

119895 119892119894)

1198881015840 (119899119896 119892119894)

when 1198881015840 (119899119896 119892119894) gt 0

(6)

If 119903119894(119862119896 119899119895) gt 1 or 119903


119895




If (1198881015840(119892119894 119899119896) gt 0)(1198881015840(119892




endifendfor










119896 119892119894) lt 119888th


119896






endfor


5 Simulations






119894(119899119896 119899119895) gt 1))




endifendfor






VehicleGrid lane

Direction vector








1 2 3 4 5606264666870727476788082848688909294


Accu

racy

ratio

()



1 2 3 4 5

14

16

18

20

22

24

26

28

30

False

pos

itive

ratio

()







1 2 3 4 50

5

10

15

20

25

30

False

neg

ativ

e rat

io (

)




1 2 3 4 5

10

15

20

25

30

35

40

45

Ave r

esid

ual e

nerg

y ra

tio (

)








6 Conclusion



Notation




119894



119896


119896and 119892

119894


119896toward

119892119894


119896

toward 119892119894


119896


119896

toward 119892119894


119895with 119862

119896

toward 119892119894



Acknowledgments


References














RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Localinvoker

Invoker Noninvoker

Clustergeneration

Runtimedecision

Clustergeneration

Runtimedecision

Joinrejectterminate





Zoom in

g1 g2 g3 g4

g4

g5 g6 g7 g8

g8

g9 g10 g11 g12







119892119894as 1198881015840(119878

119896 119892119894) In (4) 1 minus 1198881015840(119899





1198881015840(119878119896 119892119894) = 1 minus Π

119899119895isin119878119896

(1 minus 1198881015840(119899119895 119892119894)) (4)







1199051to end time 119905



119896


119894(119862119896 119899119895) which is defined





119895toward 119892

119894

during T

119903119894(119862119896 119899119895) =

1198881015840 (119862119896cup 119899119895 119892119894)

1198881015840 (119862119896 119892119894)

when 1198881015840 (119862119896 119892119894) gt 0 (5)


119895with 119899


119894is denoted as

119903119894(119899119896 119899119895) as in

119903119894(119899119896 119899119895) =

1198881015840 (119899119896 cup 119899

119895 119892119894)

1198881015840 (119899119896 119892119894)

when 1198881015840 (119899119896 119892119894) gt 0

(6)

If 119903119894(119862119896 119899119895) gt 1 or 119903


119895




If (1198881015840(119892119894 119899119896) gt 0)(1198881015840(119892




endifendfor










119896 119892119894) lt 119888th


119896






endfor


5 Simulations






119894(119899119896 119899119895) gt 1))




endifendfor






VehicleGrid lane

Direction vector








1 2 3 4 5606264666870727476788082848688909294


Accu

racy

ratio

()



1 2 3 4 5

14

16

18

20

22

24

26

28

30

False

pos

itive

ratio

()







1 2 3 4 50

5

10

15

20

25

30

False

neg

ativ

e rat

io (

)




1 2 3 4 5

10

15

20

25

30

35

40

45

Ave r

esid

ual e

nerg

y ra

tio (

)








6 Conclusion



Notation




119894



119896


119896and 119892

119894


119896toward

119892119894


119896

toward 119892119894


119896


119896

toward 119892119894


119895with 119862

119896

toward 119892119894



Acknowledgments


References














RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











If (1198881015840(119892119894 119899119896) gt 0)(1198881015840(119892




endifendfor










119896 119892119894) lt 119888th


119896






endfor


5 Simulations






119894(119899119896 119899119895) gt 1))




endifendfor






VehicleGrid lane

Direction vector








1 2 3 4 5606264666870727476788082848688909294


Accu

racy

ratio

()



1 2 3 4 5

14

16

18

20

22

24

26

28

30

False

pos

itive

ratio

()







1 2 3 4 50

5

10

15

20

25

30

False

neg

ativ

e rat

io (

)




1 2 3 4 5

10

15

20

25

30

35

40

45

Ave r

esid

ual e

nerg

y ra

tio (

)








6 Conclusion



Notation




119894



119896


119896and 119892

119894


119896toward

119892119894


119896

toward 119892119894


119896


119896

toward 119892119894


119895with 119862

119896

toward 119892119894



Acknowledgments


References














RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












119894(119899119896 119899119895) gt 1))




endifendfor






VehicleGrid lane

Direction vector








1 2 3 4 5606264666870727476788082848688909294


Accu

racy

ratio

()



1 2 3 4 5

14

16

18

20

22

24

26

28

30

False

pos

itive

ratio

()







1 2 3 4 50

5

10

15

20

25

30

False

neg

ativ

e rat

io (

)




1 2 3 4 5

10

15

20

25

30

35

40

45

Ave r

esid

ual e

nerg

y ra

tio (

)








6 Conclusion



Notation




119894



119896


119896and 119892

119894


119896toward

119892119894


119896

toward 119892119894


119896


119896

toward 119892119894


119895with 119862

119896

toward 119892119894



Acknowledgments


References














RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












1 2 3 4 5606264666870727476788082848688909294


Accu

racy

ratio

()



1 2 3 4 5

14

16

18

20

22

24

26

28

30

False

pos

itive

ratio

()







1 2 3 4 50

5

10

15

20

25

30

False

neg

ativ

e rat

io (

)




1 2 3 4 5

10

15

20

25

30

35

40

45

Ave r

esid

ual e

nerg

y ra

tio (

)








6 Conclusion



Notation




119894



119896


119896and 119892

119894


119896toward

119892119894


119896

toward 119892119894


119896


119896

toward 119892119894


119895with 119862

119896

toward 119892119894



Acknowledgments


References














RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











6 Conclusion



Notation




119894



119896


119896and 119892

119894


119896toward

119892119894


119896

toward 119892119894


119896


119896

toward 119892119894


119895with 119862

119896

toward 119892119894



Acknowledgments


References














RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Documents

Research Article Adaptive Collaborative Detection for ...downloads.hindawi.com/journals/ijdsn/2014/275609.pdfdetection focuses on the retrieve of useful data among vehicles. In this