ANETs Beacon or not to Beacon Roberta Fracchia Michela - PDF document

VANETs:ToBeaconornottoBeacon?RobertaFracchia,MichelaMeo,DarioRossiPolitecnicodiTorino,Italyemail:name.surname
@polito.itAbstract—Weaddressthebroadcastprobleminintervehicularnetworksbyconsideringtwoantipodeanapproaches:therstonemakesuseofinstantaneousinformationonvehiclesposition,whiletheotheronereliesonitslonger-termknowledge,gainedthroughabeaconingprocedure.Usingarealisticmicroscopicmodeltorepresentthevehiculartrafcow,weinvestigateandcomparetheperformanceoftheproposedalgorithmsthroughsimulation.Weshowthatasignicantperformancetradeoffexists:indeed,thoughtheapproachbasedonlonger-termknowledgeprovestobemoreefcientintermsofthechannelutilization,theinstantaneousstrategyisintrinsicallymorerobusttoerrorsduetothewirelesschannel.Finally,weshowthatthisfundamentaltradeoffcanbeturnedintoanadvantage,andweinvestigatetheeffectivenessofanhybridsolutionthatcombinesthediversityoftheaboveapproaches.I.INTRODUCTIONInthelastyears,wirelesscommunicationshaveenjoyedanamazingexpansioninmanydifferentdirections–frombringingconnectivityinmanyunder-developmentareasatlowercostswithrespecttocabling,toofferinghigh-revenuesInternetconnectivityserviceson-boardofinter-continentalights.Thisevergrowingthirstforconnectionispushingforthedeploymentofcommunicationdevicesinanewarea,i.e.,theso-calledvehicularadhocnetworks(VANETs).Typically,giventheintolerablenumberofdeathsandin-juriescausedbycaraccidents,road-safetyservicesarebroughtasaparadoxicalexampleof“killerapplication”inVANETs:inthisvision,vehiclessporton-boardwirelesscommunicationfacilitiessothat,whendangeroussituationsaredetected(eitherbyspeciconboardsensordevicesorbydriversinitiative),awarningmessagecanbeautomaticallypropagatedtovehiclesthatfollowbyadoptingadhocnetworkingcapabilities.Wire-lesstechnologies,suchasDedicatedShortRangeCommunica-tion(DSRC)[1],couldbethususedtoenhancetheperceptivecapabilitiesofthedriverinaspecularwaytowhattherear-viewmirrordoes,byprovidinga“networkedspyglass”thatallowstoforeseedangersbeyondthedrivercognitivehorizon.Inter-vehicularresearch[2]–[4]isalsofueledbytheappealofanentirelynewmarketsegment,whichin-cludesgeographically-contextualizedadvertisement,entertain-mentapplicationsandservicesaimedatimprovingpassengerstravelingcomfort.Inmanycases,thenatureoftheservicesandtheunknownandvaryingidentityoftheuserswillfavortheuseofbroadcast,ratherthanunicast,communication[5]:forexample,roadworkscouldadvertisetheirpresencethroughawirelessdevicebesidetheusualtrafcsignalsandags;sensorsonboardofvehicles(orevendirectlydeployedintheconcrete)maydetecticyroadpavement;underheavyfog,thenetworkitselfcouldrevealthepresenceofotherwiseinvisibleneighbors.Finally,suburbantollboothsorintelligenttrafc-lightsinurbanenvironmentscouldreceiveandrelaytrafcorroadconditionsinformationcomingfrominfrastructurednetworks.Inthiscontext,adoptingarealisticmicroscopicmodeltorepresentthevehiculartrafc,weinvestigatetheperformanceoftwodistributedalgorithmsforbroadcastcommunicationinVANETs.Theapproachesweconsiderareorthogonalinthesensethattheyarebasedonoppositemechanisms:theBEACONLESSapproachreliesoninstantaneousinfor-mationonvehiclespositiononly,whereastheBEACONEDoneexploitsitslonger-termknowledge,maintainedthroughtheexchangeofbeaconpackets.Letusanticipatethataninterestingperformancetradeoffexists:thoughtheBEACONEDapproachisclosetotheoptimumintermsofthechannelutilization,theBEACONLESSoneisintrinsicallymorerobusttowirelesschannelerrors.Thistradeoffsuggeststhatnodeniteanduniqueanswerexiststowhetheritisbettertoimplementandexploitbeaconingornot:differentapplicationsorservicesmayndmorefavorableeitherstrategydependingonitsspecicrequirements.Luckilythough,weshowthatthisfundamentaltradeoffcanbealsoturnedintoanadvantage,andweinvestigatetheeffectivenessofanhybridsolutionthatcombinesthediversityoftheaboveapproaches.II.RELATEDWORKTheproblemofroutinginad-hocnetworkshasbeenex-tensivelystudiedinthecontextofMobileAdhocNetworks(MANET),whoseresultshavebeenthenaturalstartingpointforVANETresearchaswell.However,duetotheintrinsicdifferencesofthesenetworks,whenappliedtoVANETs,themostpromisingroutingstrategiesarenotaseffectiveasfortheMANETcontext.Moreover,althoughunicastroutingservicemayberequired,itiswellrecognizedthatmostapplicationsoftenresorttobroadcastcommunications[5].Broadcastal-gorithmscanbemainlydividedintotwogroups:namely,beacon-lessstrategiesversusbeacon-basedapproaches.TheperformanceofthesealgorithmshasbeenextensivelystudiedandcomparedintheMANETcontext[6],[7],adoptingrathersimplemobilitymodelsthatneglecttheexistenceofcorrelationamongthemobilenodes:therefore,broadcastcom-municationneedsfurtherinvestigationsintheVANETcontext–thatfeatureveryconstrainedstring-typetopologies,ontheonehand,butstronglycorrelatedmovementsandveryhighspeeds,ontheotherhand.Thebeacon-lessalgorithmsclosesttoourscanbefoundin[7]and[8],whereseveral(eitherbeacon-basedorbeacon-less)approachesdesignedforMANETsarecompared.More specically,[7]introducesauniformdistributedwaitingtime,whileourapproach,instead,actuallyincreasesthelowestwaitingtimeinordertoensurethateverynodegatherssuf-cientinformationofthesystemstatus.Moreover,theideaoflettingthedecisionprocessdependonthedistanceisalreadypresentin[8],butthedecisionprocessisthreshold-basedratherthanprobabilistic.Amongthebeacon-lessalgorithmsspecicallydesignedforinter-vehiclecommunication,wemaycite[9]–[11].In[9]onlyoneforwardingnodeisselectedbydividingtheroadportionwithinthetransmissionrangeintosegmentsandbychoosingthevehicleinthefurthernon-emptysegment;however,thisbroadcastalgorithmrequiresMAClayermodicationstotheIEEE802.11standardanddoesnotconsiderdelayconstraints.Finally,[10]introducesawaitingtime,proportionaltothedistancefromthetransmitter,decrementedbyoneunitateachidleslot,whilein[11]thewaitingtimedependsontheprogressthatthenodecanprovidetowardsthedestination.Theuseofbeaconstodiscoverandmaintainneighborrelationshipsisgainingpopularityinbothunicastandbroad-castinter-vehicularcommunications.Inprinciple,thebea-coninginformationcouldbeusedeitherinatopology-oraposition-basedfashion.Recentwork[12]though,highlightedthattopology-basedapproaches,whileveryappealingforMANETs,arenotsuitableforVANETs:indeed,thehighvehiclemobilitymakesthetopologydiscoverandmaintenancetaskveryexpensive,introducinganexcessiveamountofcontrolmessageoverhead.Therefore,approachesthatrelyontheknowledgeofthegeographicalpositionofnodesinthenetwork,gatheredthroughGlobalPositioningSystem(GPS),aregrowingconsensusinbothacademicandcorporateresearch:thebasicideaistoexploitmorepreciseinformation(i.e.,position,directionandspeed)whileavoidingtomaintaincomplexglobaltopologicalinformation.Thisclasscanbemainlydividedingeographicforwarding[13],whereeachnodeforwardsincomingpacketstoexactlyoneofitsneigh-bors,andrestrictedooding[14],whereoodingisrestrictedtoagivenregion.Thebroadcastbeacon-basedalgorithmsdesignedforVANETsandclosesttooursare[15],[16].Aroutingalgorithmforurban-likeenvironmentsispresentedin[15]:everynodedecidestoforwardthemessageonlyifitestimatestobeclosertothetrajectorytowardthedestinationwithrespecttoitsneighbors.In[16]therebroadcastdecisionisdelayedofanamountoftimeproportionaltothedistancefromthetransmitter.Asoppositetotraditionalbeaconinginvehicularcommunications,whereeachnodeadvertisesonlyitsspeedandposition,theapproachrequireseachvehicletoadvertiseitscompleteneighborlist,generatingthusaremarkableamountofnetworktrafc:ourapproach,instead,istoseekforamechanismthatcanbeeasilyimplemented,whileprovidingatthesametimesatisfactoryperformanceevenunderrealistictrafc.III.THENETWORKSCENARIOA.ScenarioAssumptionsInordertoevaluatetheeffectivenessofthebroadcastcommunication,weconsiderabroadcastmessagepropagationnxxBPx1txxk+1xd(k+1,k)Rkx�d(k-1,k)0xrxExoutVehicles movementxinBroadcast propagationk-1xFig.1.Schematicrepresentationofthebroadcastpropagationinahighway-likescenario.Broadcastpropagationmaystartfromanytravelingvehicleorfromstationaryroad-sideunits,andistargetedtoallvehicleswithinarelevantarea.Vehiclesoutsidethisarea,instead,neverrelaythemessage,sothatthemediumremainsavailableforotherpossiblecommunicationservices.Inordertodescribethealgorithms,weadoptthenotationsketchedinFigure1.Lettheroadberepresentedbyanx-axisinthedirectionofvehiclesmovement:therelevantareastartsintheBroadcastPoint(),andcomprisesvehiclespositionedalongthex-axisat\n \r\n  .Thebroadcastmessageisforwardedovertherelevantareaexploitingmulti-hopadhoccommunications:themessagepropagatesfromintheoppositedirectionwithrespecttothevehiclesmovement(i.e.,towarddecreasingvaluesof)and,hopefully,itshouldreachallnodesuptothe-thonethatisthelastoneintherelevantarea.Letthedistancebetweennodesandbedenotedby\r !"\r#%$'&,where,clearly,thedistanceispositiveifnodeiscloserto(thannode.Theassumptionoftheroadlinearitybaresadditionaldiscussion:indeed,itcouldbearguedthatroadsareactuallyquiteconvolutedand,infact,thegeographicaldistance)$+*,.-*/ canbesmallerthan\r0$1*, forsome.However,wepointoutthat,beingvehiclesequippedwithGPS,theyarealsolikelyequippedwithanavigationsystemaswell:therefore,adigitalmapwillbeavailableatthereceiver.Inthiscase,thepositionadvertisedbytransmittersinthebroadcastpacketscanbere-mappedtoa“linearized”roadportion,andthustheactualroad-distance,ratherthantheair-distance,couldbeeasilyconsidered.Moreover,suchroadwindingcanbeexpectedtooccupyarelativelysmallhighwayportioncomparedtotheroughlystraightone.Finally,thestudyofalinearroadstretchisapreliminarybutnecessarystep,beforemorerealisticbutcomplexscenarios(e.g.,involvingintersection,motorwayoverpassesandhighwayjunctions)couldbetakenintoaccount.Anotherimportantremarkisthat,althoughbroadcastpack-etslikelyrefertoasingletrafcdirection,theyareneverthelessreceivedbybothdirectionsofthetrafcow.However,forthesakeofsimplicity,inthefollowingweconsidertheroadstretchinthebroadcastpropagationdirectiononly.First,wepointoutthatthisisbasicallyequivalenttoassumethatvehiclestravelingintheoppositedirectioncandiscriminate,viatheGPS,ifthereceivedmessageispertinenttothedirectionoflanestheyaretravelingalong(whichcanbedonesimplybytestingwhetherthedirectionofthebroadcastmessage propagationistheoppositewithrespecttotheirtravelingdirection).Second,westressthatvehiclestravelingintheoppositedirectioncouldrelaythenon-pertinentmessagesonpurpose:indeed,inlowdensitynetworks,thesevehiclescouldhelpfullyextendthenetworkconnectivityandthemessagecoverage–althoughthismaybechallengingduetohighrelativespeedsamongvehicles,itmayaswellrepresentanadvantagetoexploit.B.CommunicationAssumptionsLetusnowelaborateonnodescommunicationcapabilities.Wedonotfocusonaspecictechnology,butweassumethattheon-boarddevicetransmitsatadatarateof2MbpsonaR=200mtransmissionrange,wherethelowdata-ratehasbeenchosenbyvirtueofitsbetterresiliencetothewirelesschannelerrors.Also,wedonotinvestigatethecontentofthebroadcastmessagebutweassumethati)thepositionofthebroadcastinitiator,andii)thepositionofthelastrelayvehiclearereported,sothatmessagerebroadcastcanbeeasilyconnedintherelevantarea.Besides,weassumethatpacketheadercontainstheiii)broadcastinitiatoridentieraswellasiv)arandomlychosenpacketidentier,assignedoncebytheoriginalsource:allnodesarerequiredtocachethisinformationonasoft-statetableand,ateveryforwardinghopinthenetwork,eachnodeperformsatablelookup,avoidingtorebroadcastamessagecarryingthesameidentiersmorethanonce.ForwhatconcernstheMACprotocol,nodesadopta0-persistentCarrierSenseMultipleAccess(CSMA)mechanism:inordertoavoidcollisions,beforestartingatransmission,theysensewhetherthechannelisbusy.Inthelattercase,themessagetransmissionisdelayed,foranamountoftimeslots&#x;000;23547698:suniformlydistributedbetweenzeroandthecontentionwindowsizeA)BCEDF*,untilthemediumissensedidle.Finally,inthecasewhereaposition-basedroutinglayerisimplementedbythecommunicationdevice,weadoptthecommonbeaconformatandthestandardbeaconingproceduredescribedintheliterature.Tobemoreprecise,beaconsareusuallyabout20Byteslongandcarryinformationregardingi)thevehicleidentier,ii)itsgeographicalpositionandiii)itsspeed.Everyvehiclecachesthebeaconinformationalongwiththetimeofitsreception,whichwillbelaterusedtoestimatetheneighbor'sposition.Theinter-beacontransmissionintervalisusuallyaxedvaluebetween1sand5s.Moreover,inordertoavoidsynchronizationandbeaconcollisions,wejitterthetransmissionofeachbeaconasin[13],sothattheinter-beacontransmissiontimeisuniformlydistributedin[0.5,1.5].Wepointoutthatthebeaconstransmissionrequiresaverylowchannelutilization.Indeed,considering20-Bytesbeaconsandequalto1s,thebeaconingtaskofeveryvehicleconsumes160bpsofphysicallayerbandwidth:atthehighestdensityweconsider,namely50veh/Km,the20vehicleslyinginthesametransmissionrangewouldthususe3.2Kbps.Givenaphysicallayerchannelcapacityof2Mbps,thechannelutilizationwouldbethusaslowas0.16%;similarly,wecanderiveachannelutilizationof0.03%whenGHs.C.VehicularTrafcModelsTheperformanceofwirelesscommunicationalgorithmsinmobileadhocnetworksstronglydependsontheadoptedmobilitymodel,especiallywhen,asinourcase,ahighmobilityscenarioisanalyzed.In[17],byconsideringanumberofpopulartrafcmodels,weshowedthat,althoughnetworkconnectivityremainslargelyunaffectedbythespecicusedmodel,thetrafcmodelplaysacriticalroleindeterminingtheperformanceofinter-vehicularcommunicationalgorithmsunderstudy.InthispaperweadoptarealisticmicroscopictrafcmodelthatfallsinthecategoryofCoupledMaps(CM)models.Thesemodelshavecoarse-graineddiscretetime,whilespaceiscontinue,and,astestiedbyindependentempiricaltrafcmeasurements[18],[19],theydisplaypropertiessimilartotherealtrafcdynamics(consideringbothindividualvehi-clemovementsandthecorrelationbetweenthebehaviorofneighboringvehicles).ThemostpopularCMmodelisduetoGipps[20]but,inreasonofitssimplernotation,wereportheretheformulationofKrauß's[21].EachvehicleisindividuallyrepresentedbyastatevectorIKJ describingitsspatiallocation,speedanddistancefromthevehicleahead,and,ateachtimeslot,thestatevectorofeachvehicleisupdatedaccordingtothefollowingsetofrules:LMMMMMMNMMMMMMOVelocity-update&#x;000;J39PRQTSUVJ\n-XWY\rZZ[\JWS93V]^_%`Ja-cb\rdJ39PQ/SdJfePhgjiJkV]blI`/=KJWS93$mbonpiMotion-updateqrs-tJwherebdenotesthemaximumvehicleacceleration,uthemaximumdeceleration,nisthenoiseamplitudeandpisarandomnumberinv=F*w.Moreover,ifxJisthevelocityofthecaraheadattimey,wefurtherindicatewithJkJz-mxJ K{f;theaverageofthevehiclesvelocitiesusedtodeterminethesafespeedboundJ,39PRQTS.Therstrulesdescribedeterministiccar-followingbehavior:driverstrytoaccelerateexceptwhenthegapfromthevehicleaheadistoosmallorwhenthemaximumspeedisreached.Vehiclesspeedislimitednotonlybythemaximumvelocity&#x;000;JePRgbutalsobythedesiredaccelerationbandbythesafespeedboundJ3dPRQTS,whichdependsonthelocalconditionsofthetrafc.Thespeedisalsosubjecttoarandomperturbation:withaprobabilityp,avehicleendsupbeingslowerthanwhatcalculateddeterministically:thisparametersimultaneouslymodelseffectsofi)speeductuationsatfreedriving,ii)over-reactionsatbrakingandcar-following,andiii)randomnessduringaccelerationperiods.Inoursimulations,wesetJePhgto135km/hrandthetime-stepgranularityisy|*s,eveniftheseunitsareassumedimplicitlyandleftoutoftheequations.Theparametersaresettob\ruTnh }=~;jK=F~€F*,~=l ,valuesthatarestandardlyusedintheliterature[21].IV.BROADCASTALGORITHMSInthissectionwedescribethetwoalgorithmsthatweusetocompareBEACONEDandBEACONLESSbroadcastcommu-nications.Forcomparisonpurposes,inwhatfollowswewillalsoconsidertheoptimumcentralizedsolutionthatselectsthe Eval Coverage txReceive new noyesx c~Avoid torebroadcastwith 0-p CSMAForward pktUpdate Neighbor Select Closest~cFig.2.Flow-chartsoftheBEACONEDalgorithmsminimumrelynodessetthatmaximizesthemessagecoverage,aswellastheverysimple–oodingstrategy,rulingthat,uponthereceptionofanewmessage,eachnodeforwardsthepacketwithindependentprobability.A.BeaconedAlgorithmThebasicideaoftheBEACONEDalgorithmisthat,when-eversomeinformationthathastobebroadcastedisreceived,eachnodeestimatesthepositionofitsneighborsbyexploitingtheinformationpreviouslyexchangedbyroutinglayerthroughtheuseofbeaconpackets:theneighbors'positionestimateisthenusedtodecidewhetheramessageshouldbeforwardedornot.Theforwardingdecisionaimsattrading-offthenumberoftransmittedmessagesandtheprobabilitytoinformtheve-hicles:theoptimaltrade-offisderivedwhenonlythefurthestawayinformedvehicleforwardsthemessage.Weassumethattheroutinglayermaintainsneighborhoodstateinformations,calledtheneighborsset‚recordingforeachneighborthetuplefƒy#ƒ#ƒJ# whichstoresthenodeidentier,thetimeofthebeacontransmissionƒy#,thepositionƒ#andthespeedƒJ#ofthenodeatthebeacontransmissiontime.Moreover,neighbortableentriesaremaintainedbasedontheirageusingatimer2,whichcanbeconsideredasan“agethreshold”:belowthethreshold2,therouting-layerinforma-tionisconsideredtobeup-to-date,whereasanymemberofthe‚setolderthan2isdiscarded.Bypreliminarysimulations,weveriedthat,asthechannelutilizationduetothebeaconingisverylow,theuseofajitteredbeacontransmissionkeepscollisionsofbeaconmessagesnegligible.Asaconsequence,whenthewirelesschanneliserror-free,ineverybeaconinginterval,nodesgatheracompleteinformationoftheirneighborhood.However,inordertoincreasetherobustnessincaseoflossywirelessmedia,weselecttheagethresholdtobe&#x C+E;&#x?000;2„…;,thusconsideringvalidtheinformationreceivedduringthelastorthesecond-lastbeaconingcycle.ConsideringthenodesofFigure1,letusassumethatattimeynode†Treceivesamessagefromnodey5.There-broadcastingdecisionprocessat†/,sketchedintheow-chartofFigure2,initiallyevaluatesthebroadcastmessagecoverageareaasthezonefrom\r8‡gtoF8‡g$ˆ;notethatthisinformationisprecise,sincetheactualtransmitterpo-sition8‡gispiggybackedinthebroadcastpacketheader.Then,†Tupdatestheestimateofitsneighborsposition,byusingtheinformationstoredonitsneighborset‚z‰gas&#x C+E;&#x?000;Rx m0copies? Listen forwaitTu() min u=U(0,1) Extractd Updatemind min&#xP d ;&#x ;&#x ?0;0 ?noyesAvoid torebroadcastnoyesReceive new with 0-p CSMAForward pktyesnoFig.3.Flow-chartsoftheBEACONLESSalgorithmsx#zƒ\r#-y%$ƒy9#Š ƒJf#,‹\rŒ:‚‰g.Thereceiverthenindividu-atesthetheclosestfollowerŽamongitsneighbors:Žistheclosestneighborinthemessagepropagationdirection,thus&#xP d ;&#x ;&#x ?0;&#xP d ;&#x ;&#x ?0;ŽV‘K’}“s”–•F#˜—f™š‰g!$x\r#| =l .Finally,†Tdecidestorebroadcastthemessagewheneitheri)Žisoutsidethecoveragerange(i.e.,Ž›œT\ryinFigure1),orii)Žisinsidethecoveragerangebutwithinadistancežfromthecoverageborder(i.e.,ŽŸŠ).Theabovetworules,separatelydescribedforthesakeofclarity,canbeexpressedinasinglethreshold-baseddecision:thus,themessageisforwardediff,x\r ¡8‡g$¢ˆ-cž.Observethaterrorsontheneighborspositionestimationhavenegativeimpactontheperformance;inparticular,thesignoftheerrorleadstooppositeconsequences:over-estimatingthedistanceofaneighborraisesthenumberofunnecessaryrelaynodes,whereasdistanceunder-estimationreducesthenumberofreachedvehicles,thusincreasingtheprobabilityofaprematureinterruptionofthebroadcastprop-agation.Therefore,thesafetymarginžisincludedinordertoreducethe“missedforwarding”errors(i.e.,nodesthaterroneouslydecidetoavoidrebroadcasting)inthedecisionprocess,inordertoaccountfor:i)under-estimationoftheneighborpositionandii)possiblymissingneighborhoodin-formationduetobeaconloss.B.BeaconlessAlgorithmThesecondapproachwepropose,orthogonaltothepreviousone,doesnotrelyonanytopologicalinformationofthenet-work:everynode,beingunawareofothervehicles'position,shoulddecideindependently,andthusprobabilistically.TheBEACONLESSapproachisbasedontwomainideas:i)tobasetheforwardingdecisiononthedistancefromtheclosestneighboringrelaynode,andii)tointroduceashortwaitingtimebeforethemessageforwarding.Therationalebehindtheideaoflettingthedecisiondependonthedistanceisthefollowing:whenanodehearsthemessageforthersttimeanditsdistancefromthetransmittingnodeissmall,thentheadditionalcoveragethatcanbeachievedbyre-broadcastingthemessageisalsosmall1:thus,thedecisiontoforwardthemessageshouldbetakenwithlowprobability.Theroleofthewaitingtimeistoallownodestolistenfornewcopiesofthebroadcastmessage:thisyieldstoabetterestimateof1Thisassumptionisparticularlyrelevantifvehicleshavethesametrans-missionrangeandinthecaseofunidirectionalpropagation. thesmallerrelaydistance,whichisthenusedtotunetheforwardingdecisionprocess.TheBEACONLESSscheme,whoseow-chartispresentedinFigure3,worksasfollows.Assoonasanode†Thasreceivedforthersttimeabroadcastmessagefromnodey5,thenodesetsthevariablee#tothedistance\ry5@d†T\r andstartssensingthechannelforatime2£Ph#78,duringwhichthenodechecksifothercopiesofthesamemessagearereceived.Thewaitingtime2£Ph#¤8issetproportionaltothetime28‡gnecessarytotransmitafull-siwemessageaugmentedbyafactor¥plusarandomamountuniformlydistributedinv=A)Bwof2@3_4–698longtime-slots:2£PR#¤8|V¥\r2'8‡g¦-§¨v=FA)Bw©2ª3547698(1)Noticethat2ª£Ph#78iscomposedbyaxedandavariablecomponent.Theroleofthexedcomponent¥\r28‡gistoensurethatnode†/canacknowledgeifothervehiclesareforwardingthemessage,andthusdecidetorefrainfromrebroadcast2.Theroleofthevariablecomponent§sv=A)Bw–23_476d8istwofold:rst,itavoidsthesynchronizationofretransmissionsfromnodesthatdecidedtorebroadcastthemessage;second,andmostimportant,itallowstheunorderedretransmissionofnodesbelongingtothesametransmissionrange.Assumethat,aftertime2£Ph#78,node†Thascollected]copiesofthemessagefrom]nodes~~~de.IfatleastonenodeisfurtherawayfromtheBP,i.e.,itexistsanode#suchthat\rI#dK†/' «¡¬=,then†Tcansafelyavoidtoforwardthemessage:indeed,themessagehasalreadycoveredanareawhichisoutsidethetransmissionrangeof†Tandare-broadcastfrom†Twouldbeuseless.Otherwise,ifallthetransmittingnodesareclosertoBPthan†T,i.e.,if‹\r\r#d†T' ­®=,then†Tcomputestheminimumestimateddistancee#sq¯“k”–•#`e#'#K†/' iandentersanewwaitingphase.Conversely,whenduringthesensingperiodnocopiesofthemessagearereceived,themessageisthenforwardedwithaprobability˜e#F ,increasingwiththeminimumestimateddistancee#fromtherelayinghosts.Thus,inthecasewherethenodedecidestorebroadcast,thebroadcastmessageisdeliveredtotheMAClayer.Amongtheseveralfunctionfamiliesthatcanbeusedtosettheprobability°e# ,wechoose:°e# «*U$*¦$e#{fˆ\n d±k³²´ µ*(2)Notethatinthesupporte#¶Œ·v=FKˆaw,thecurvesaremonotonicallyincreasinginbothe#and²,andthatthefunctiondegeneratesintoe#{fˆwhen²¸…*.Theimpactof²in(2)isthoroughlydiscussedinthenextsection.V.PERFORMANCEEVALUATIONResultsreportedinthissectionareobtainedwithadis-creteeventsimulatorthataccuratelydescribesthescenariopresentedinSectionIII-B.Inparticular,thesimulatorproperlydescribesnetworkingdynamicslikecollisions,propagation2Sinceweveriedbysimulationthattheperformanceresultsareonlymarginallyaffectedbyanyvalueof¹.º0»,inthefollowingweset¹U¼0½. 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 10 20 30 40 50Accuracy PinfoVehicular Density r [veh/km] 0.85 0.9 0.95 1 Reliability 10 20 30 40 50 0 2 4 6 8 10 12 14 16Overhead o OptimumB=1sB=3sB=5sFig.4.BEACONEDsystemperformance:beaconinginterval¾impactdelays,transmissiontimes,CSMAdelaysandimplementsthevehiculartrafcmodeldescribedinSectionIII-C.Theperformancemetricsweconsiderare:i)theaccuracyofthemessagepropagation,denedasthepercentageofvehiclesreachedbythebroadcastmessage,thatisalsotheprobabilitythatavehicleisinformed;ii)thereliability,denedastheratiooftheaccuracywithrespecttoareference(e.g.,optimum)value;iii)thepropagationoverhead,expressedintermsofthenumberoftransmittedmessages,iv)theredundancy,denedastheratiooftheoverheadwithrespecttoareference(e.g.,optimum)value,andv)thetimelinessexpressedasthedelaybywhichthelastinformedvehiclereceivesthemessage.Theperformanceindicesareobtainedwithacondencelevelof99%andacondenceintervalof2%,consideringa2kmlongrelevantareaoutofa8kmlongsimulatedroadstretch.Inthefollowing,whenwetalkaboutoptimumstrategywerefertothecaseinwhichtheonlynodethatrelaysamessageisthefurthestnodereceivingthemessage.Thisstrategy,whenthewirelesschanneliserror-free,istheoptimalone,sinceitjointlyminimizesthenumberoftransmittedmessagesandmaximizesthenumberofinformedvehicles.Asaresults,onalineartopology,alsothenumberofcollidedmessagesisminimized.IdealWirelessChannelInthissectionweevaluatetheperformanceoftheproposedalgorithms,comparedtothatoftheoptimalstrategy,whenthewirelessmediumiserrorfree.LetusrstconsidertheBEACONEDapproach.Figure4depictsthealgorithmperformanceasafunctionofthedensity¿andfordifferentvaluesofthebeaconingintervalÀŒ`,*,KDFHis;initially,wedonotconsideranyerrormarginonthepositionestimate,thuswesetžÁ¯=m.TheveryleftplotofFigure4reportsthealgorithmaccuracyinfo,thatisthepercentageofinformedvehicles.Theoptimumcentralizedstrategyisalsoreportedasareference.Itisinterestingtonoticethatwhenvehiclesareinfree-ow(¿¡";f=veh/km),BEACONEDperformanceisveryclosetothebestcase–whichismainlydrivenbythenetworkconnectivity–foranyoftheconsideredvaluesof:indeed,sincetheroadisuncongested,vehiclesunlikelymodifytheirspeedin5secondsorlessand Reliability RatioVersus Optimum0.900.951.00Versus B=1s, E=0mE=8mE=2mE=0m 10 20 30 40 50Redundancy RatioVehicular Density r [veh/km] 10 20 30 40 500.901.001.101.20 Fig.5.BEACONEDsystemperformance:errormarginÂimpactthepositionestimateisthusveryaccurate.Onthecontrary,whentrafcisjammed,vehiclesspeedmayrapidlychangeandtheaccuracyofthepositionestimateispossiblysignicantlythreatenedasincreases.Theinsetplotdepictstherelativesystemreliability,thatistherelativeaccuracywithrespecttotheoptimumstrategy.In-terestingly,reliabilityshowsanegativepeakaroundthecriticaldensityirrespectivelyofthebeaconingfrequency.However,forhigherdensities,asmallerbeaconingintervaltranslatesintohighersystemreliability,whereasforlargevaluesofthereliabilitydecreasesas¿increasesabovethecriticalthreshold.Finally,therightplotofFigure4reportsthealgorithmoverheadÃ,expressedastheamountoftrafcgeneratedintherelevantareaintermsofthenumberoftransmittedmessages.Notethatwedonotincludethebeaconingmessagesoverhead,reportedinSectionIII-Bintermsofchannelutilization,thusdirectlyinvestigatingtheoverheadduetothebroadcastmessagepropagation.NoticethatÃalsocorrespondstothenumberofrelaynodes,sinceeveryvehicleforwardsthemessageatmostonetime:itcanbeseenthat,whenÄ1*sorGDs,thenumberofrelaynodesrunningthisdistributedalgorithmonlyslightlyexceedstheoptimumoneobtainedbyacentralizedstrategy.WhenrÅHs,asystematicdistanceunder-estimationleadsnodestorefrainfromrebroadcasting(i.e.,aloweroverhead)whichinturnsreducesthealgorithmeffectiveness(i.e.,aloweraccuracy).Inordertoreducethebeaconingoverheadandamelioratethesystemreliabilityaswell,letusnowintroduceanerrormarginžonthepositionestimate:intuitively,thehighertheadoptederrormarginis,thehighertheaccuracyis.Consider-ingtheworstcase„«Hs,weanalyzethebenetsofintro-ducingtheerrormarginž,byconsideringžim.Figure5depictsthereliability(top)andredundancy(bottom)ratioswithrespecttotheoptimumstrategy(left)andwithrespecttothecasewithÀ·*sandžÇÈ=m(right).Observingthetoprightplot,itcanbegatheredthattheintroductionofanerrormarginž…1Æmmakesthe"³Hssystemmorereliablewithrespecttosystemswithhigherbeaconingfrequencyandwithouterrormargin(i.e.,„X*s,žÉ¶=m).Also,consideringthebottomrightplot,thesebenetscomeattheexpenseofamodest10%increaseonthe 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 10 20 30 40 50Accuracy PinfoVehicular Density r [veh/km] 0.5 0.6 0.7 0.8 0.9 1 Reliability 10 20 30 40 50 0 5 10 15 20 25 30 35 40Overhead o OptimumK=8K=2K=1Fig.6.BEACONLESSsystemperformance:exponentÊimpactrelayredundancy:inotherwords,onlyabouttwoadditionalmessagesperbroadcastcycleareneeded,butatthesametimetheperiodicbeaconingoverheadisreducedbyafactorof5.Observingtheleftplots,itcanbeseenthattheintroductionofanerrormarginËÈ̬ÍmbringstheÎÏÌÑÐssystemperformanceveryclosetotheoptimumintermsofreliability,attheexpenseofaredundancyincreaselowerthan25%.LetusnowconsidertheperformanceoftheBEACONLESSapproachwhenthewirelessmediumiserrorfree.Figure6depictstheresultsfordifferentvaluesofÒin(2)asfunctionofthevehiculartrafcdensity.Moreondetails,thegurereportstheaccuracy(leftplot),thereliability(insetplot),andtheoverhead(rightplot),achievedbytheBEACONLESSapproachandbytheoptimumcentralizedstrategy,usedagainasareference.Firstofall,noticethat,forhighvaluesofÒ,reliabilitycanbemadearbitrarilyclosetotheoneoftheoptimumstrategy,evenaroundthecriticaldensityregionwheretheBEACONEDalgorithmhassignicantperformanceproblems.Moreover,unlikeintheBEACONEDcase,reliabilityincreaseswithvehiculardensity,becausetheprobabilitythatatleastonenodeforwardsthemessageincreases.Finally,theoverhead,thatalsoincreasesasthevehiculardensityincreases,intheÒCÌÓÍcaseisatmostthreetimestheoneoftheoptimumcase.Forwhatconcernsthedelay,showninTableI,itisworthnoticingthat,astheoveralldistancecoveredbythebroadcastmessageincreases(Ô¨Õ20veh/Km),theaveragedelayincurredbythemessageincreasesaswell.Athigherdensities,theBEACONLESSdelayfurtherincreasesbecausethenumberofprobabilisticallychosenrelayincreasesandthustheaveragetimetoaccessthemedium.Conversely,theBEACONEDdelay,accordinglywiththeoptimumone,slightlyreduces,sincethenumberofnecessaryrelaynodes,thatarethefurthestvehiclesinthecoveragearea,reducesasthedensityarises.B.Non-idealWirelessChannelTheprevioussectionhashighlightedsomeimportantdif-ferencesbetweentheBEACONEDandtheBEACONLESSapproach:theformeristheleastredundantandthemostperformingintermsofdeliveringtime,whilethelatteristhemostreliable.However,thesedifferencesaremarginal, TABLEIOPTIMUM,BEACONLESSANDBEACONEDAVERAGEDELAY[MS]DensityOptimumBEACONEDBEACONLESS[veh/Km]E=0mE=8mK=1K=85394138841312058596021430350475450271419sincebytuningthealgorithmparameters,theBEACONEDapproachcanbemademorereliable(e.g.,increasingž)andtheBEACONLESSlessredundant(e.g.,decreasing²).Inordertocompleteourcomparativeanalysisofthetwoapproacheswenowintroducemorerealisticwirelesschannelmodels,sothatwecanalsoassesstherobustnessofthetwoalgorithmstoerrorsoverthewirelesschannel.Ifnototherwisespecied,weconsidertheBEACONEDstrategywithĵHsandžXÆmandtheBEACONLESSwith²ÖÆ.Letusnowintroducetwowirelesschannelmodels:intherstone,theprobabilitythatamessageisnotcorrectlyreceivedisBernoulliwithprobability×,independentlyfromthereceiverdistancefromthetransmitter.Conversely,inthesecondmodelthelossprobabilitydependsonthedistancebetweentransmitterandreceiver:indeed,itiswellknownthattheelectromagneticsignaldegradeswiththedistance,andem-piricalstudies[22],[23]showthatatransitionalregionexists.WedenotebyØ@ˆtheamplitudeofthetransitionalregionandweassumethat:i)below*I$ÙØ% 9ˆcommunicationislossless,ii)within*Ú$ÛØz dˆandˆthepacketlossprobabilitysharplyincreasesandiii)aboveˆ,thatisoutsidethetransmissionrange,communicationwillcertainlyfail.Inthefollowing,weusuallyrefertotheaveragechannellossprobability,whichisequalto×andØzˆ!{f;fortheBernoulliandtransitionalchannelrespectively.Summarizing,@Ü%S‰˜ ÝÞ×if¨ßˆ*if ®ˆ@à\r‰P˜ ÈLNO=if¨ß«*U$¢Ø% 9ˆWY@áY'â,ã‡ä�âäif*U$mØ% 9ˆX¡¨ßˆ*if ®ˆTheabovemodels,despitetheirsimplicity,allowarstgradeinspectionofthesystembehaviorinpresenceofnon-idealwirelesschannels;besides,anadvantageofthislevelofabstractionisthatsuchmodelsarenottiedtoapeculiarphysicalmodulation,nortoaspecichardwareplatform,nortoagivenenvironment,butareapplicabletoamoregeneralextent.Also,wepointoutthatwirelesschannelmodelssuchasGilbert's[24],whichintroducetime-correlationinthewirelesschannel,arenotsuitedforoursetup:indeed,sincethemessagepropagationhappensonlyoncealongagivenchannel,theeffectofthetime-correlationislikelytobenegligible.Onthecontrary,weexpecttheeffectofthespace-correlationintroducedbythetransitionalmodeltoplayaremarkablerole.Figure7depicts,asafunctionoftheaveragelossproba-bility,theaccuracyoftheBEACONLESSandtheBEACONEDalgorithms,fortwodifferentvaluesofthevehiculardensity,namely20veh/km(leftplot)and40veh/km(rightplot).First, 0 0.1 0.2 0.3 0.4 0.5Accuracy PinfoAverage Loss ProbabilityDensity r=20 [veh/km] 0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Density r=40 [veh/km]BED: BernoulliBED: TransitionalBSS: BernoulliBSS: TransitionalFig.7.BEACONLESS(BSS)andBEACONED(BED)accuracyasafunctionofwirelesschannelerrors,forTransitionalandBernoullichannelmodels,atlow(left)andhigh(right)vehiculardensitynoticethattheBEACONLESSalgorithmisfarmorerobustwithrespecttotheBEACONEDserviceirrespectivelyofthechannelmodel.Indeed,evenintheworstcase,whenhalfofthepacketsarelostduetowirelesschannelerrors,morethanhalfofthevehiclesareeffectivelywarnedbytheBEACONLESSalgorithm,whereasthispercentagedropstolessthanonefthwhentheBEACONEDstrategyisconsidered.ThereasonisthatfortheBEACONLESSstrategypacketlosseshavetheeffectofvirtuallyreducingthedensityoftherelaynodes,and,thankstothehighredundancy,theeffectoferrorsisvisibleonlyatlowdensity(e.g.,20veh/Km)andhigherrorprobabilities.TheforwardingdecisiontakenbyBEACONEDapproachisinsteadbasedonnodespositionestimatesoastoreduceredundancy;therefore,iferrorprobabilityislargeitcaneasilytranslateintheinterruptionofthemessagepropagation.Noticealsothatweconsiderthepossibilitythatbeaconingmessagestoomaybelostduetoradioerrors,leadingtoamoreimpreciseneighborhoodknowledge.Moreover,theBEACONEDstrategyisnotonlyaffectedbythelossamount,butitisalsosensitivetotheir“placement”:indeed,sincethealgorithmisveryeffectiveinselectingonlythefartherrelaynodes,andsincethetransitionalchannelyieldshigherlossprobabilitiesatlongerdistances,theperformancedegradationismoreremarkableunderthetransitionalchannelmodel.Onthecontrary,theperformanceoftheBEACONLESSstrategyisalmostthesameforthetwochannelmodelsanditisdrivenonlybytheamountoflosses.Giventhenon-idealityofthewirelesschannel,itisnolongerpossibletodeviseanoptimumcentralizedsolutiontotheproblem.Weuseasareferencethesimpleprobabilisticå–oodingstrategy,regulatingtheaccesstothechannelwiththeæ-pCSMA:whenåÌ1çthesystemreliabilityisenforced,butatthepriceofabroadcastpacketstorm;conversely,byloweringå,toe.g.1/2,theredundancyishalvedattheexpenseoflowerreliability.Figure8depicts,foravehiculardensityofÔ´Ì·è,æveh/km,thereliabilityandredundancyoftheBEACONEDstrategy(ÎGÌÐswith˳ÌVÍmandÎXÌ«çswithˬÌéæm),theBEACONLESSand1/2–oodingapproachesnormalizedversustheç–oodingperformanceforboththeconsideredwirelesschannelmodels.Notethat,althoughtheBEACONLESSalgorithmreducestheredundancyofabout Reliability RatioTransitional Channel 0 0.2 0.4 0.6 0.8 1Bernoulli Channelflood1/2BSS(K=8)BED(B=1, E=0)BED(B=5, E=8)0.50.40.30.20.1Redundancy RatioAverage Loss Probability0.50.40.30.20.1 0 0.1 0.2 0.3 0.4 0.5Average Loss ProbabilityFig.8.BEACONLESS(BSS)andBEACONED(BED)reliability(top)andredundancy(bottom)ratiooverthe1-oodingalgorithm,forTransitional(left)andBernoulli(right)channelmodels60%–70%,itsreliabilityispracticallyindistinguishablefromç–ooding.Forinstance,comparingtheBEACONLESStothe1/2–oodingstrategy,observethatfortheBernoullichannelwithê¢ÌµæFëÐ,theBEACONLESSapproachsuccessfullyalerts28%morevehiclesusing8%lessmessages.Conversely,theBEACONEDapproachreducesofabout95%thenumberofexchangedmessages,butthesystemreliabilitydropssigni-cantly(slightlymorethan20%intheworstcase),andeventheintroductionofasafetyerrormargindoesnotmitigatetheeffectofthewirelesschannelerrors.C.ToBeaconorNotToBeacon?TheprevioussectionshighlightedthatseveraltradeoffsexistbetweenBEACONLESSandBEACONEDapproaches.However,itisnotpossibletogiveadeniteanduniqueanswertowhetheritisworthintroducingabeaconingprocedureornotintheVANETcontext:indeed,differentapplicationsorservicesmaypreferonesolutionortheother,dependingontheirspecicrequirements.Forexample,aroad-safetyserviceshouldimplementaBEACONLESSalgorithm,sinceitsintrinsicredundancyguaranteeshighreliabilityandrobustnesstowirelesschannelerror.Onthecontrary,whenthesafetyofthedriversinnolongerendangered,thesmalleroverheadoftheBEACONEDstrategymaybepreferred.Anotheroptionconsistsinproposinghybridsolutionsthattrade-offamongreliability,robustnessandoverhead.Inhybridsolutions,thedecisionofwhethertorelaythepacketistakenbycombiningthedecisionsthattheBEACONEDandBEACONLESSapproacheswouldtake.Let�Î\nì@íîmï/æFðçñand�Î.òoò®îµï/æFðçñbethebooleanrandomvariablesrepresentingtheoutputofthedecisionprocessesoftheBEACONEDandBEACONLESSalgorithms,respectively.Î)òoòissetaccordingtothefunctionó¨ô˜õoöÚ÷–øFùin(2).Inthehybridsolution,thetwodecisionsarecombinedinthefollowingway:thepacketisforwardedaccordingtothebooleanvariableÎ)ìIí+ú¢û–ô‡üÛýþù|ÿmÎòoò
,whereüisauniformrandomnumberinûæFðç.Inotherwords,theprobabilitythatthepacketisforwardedintheBEACONEDstrategyisincreasedbytheprobabilitytobeforwardedintheBEACONLESSstrategy,andthelasttermis5045403530252015105Hybridation Degree H [%]Vehicular Density r [km/hr]Hybrid Reliability Ratio0.300.500.750.901.252.002.503.00504540353025201510 0 10 20 30 40 50 60 70 80 90 Hybrid Redundancy Ratio0.200.300.400.502.004.006.008.00vs Beaconedvs BeaconlessFig.9.HybridatingtheBEACONLESSandBEACONEDapproaches:Relia-bility,redundancycontourplotasafunctionofthehybridationdegreeandthevehiculardensityweightedbyafactorþthatrepresentsthehybridationdegree.Werunseveralsimulations,consideringdifferentsettingsoftheBEACONEDandBEACONLESSalgorithms,channelerrormodelsandprobabilities,anddifferenthybridationdegreesaswell.Inordertoquantifytheeffectivenessofthehybridsolution,wereportresultsfortheBernoullichannelwithpacketlossprobabilityêXÌæFëÐ.Weconsiderθ̛ÐsandËÌÍmfortheBEACONEDapproach,andÒ̶Ífortheprobabilisticrebroadcastfunction(2).Figure9depictsthecontourplotfortheredundancyandreliabilityratioofthehybridapproachoverboththepure-BEACONEDandthepure-BEACONLESSalgorithms,asafunctionofthevehiculardensityÔandofthehybridationdegreeþ.Forexample,letusfocusonreliability(leftplot).Asexpected,thehybridapproachallowstogreatlyenhancereliabilityofBEACONED(ofevenafactorof3),inawaywhichisfarmoreeffectivethanwhatwasprovidedbytheintroductionoftheerrormargin.WhenÔ�æ,itispossibletodoubletheBEACONEDeffectiveness(linelabeledwith2.0fortheBEACONED)whileachievinghalfoftheBEACONLESSeffectiveness(linelabeledwith0.5fortheBEACONLESS).ThecurvesdivergeforÔý�æ,whereahigherhybridationdegreeisrequiredtokeepafactor2gainovertheBEACONEDapproach.ThehybridationrequiredtoachievehalfofBEACONLESSreliability,instead,sharplydecreasesasthevehiculardensitydecreases,becauseforsparsernetworksconnectivityandwirelesserrorsdominatetheperformanceofthealgorithms.Figure9alsosuggeststhatthehybridationdegreeþshouldnotbekeptconstantoverallvehiculardensities,ratheritshouldbeadaptivelyadjustedtothetrafcconditions.Asanongoingwork,weareexploringthepossibilitytoestimatethevehicledensitythroughthebeaconingproceduresoastoadaptivelysetthevaluesofthehybridstrategyparameters:intuitively,whenevertheestimatedvehiculardensityislow,thehybridationdegreeshouldbehigherandvice-versa.Finally,Figure10depictsthedelayatthefarmostinformednodeasafunctionofthevehiculardensityfortheBernoullichannelwithê«ÌæëÐfordifferenthybridationpercentages. 20 40 60 80 100 120 140 5 10 15 20 25 30 35 40 45 50Delay at Maximum Distance [ms]Density r [veh/km]H=75%H=50%H=25%(BED) H=0%Fig.10.HybridatingtheBEACONLESSandBEACONEDapproaches:Delayatmaximumdistance,fordifferenthybridationdegrees,asafunctionofthevehiculardensity ThepicturenotonlyconrmsthesmallBEACONEDdelay,butitalsoshowsthathybridationprovidesameanforsmoothlytuningperformancebetweenthetwoorthogonalalgorithms.Also,itshouldbekeptinmindthatthedelayperformanceisveryeffective,sincevehiclesattheendoftherelevantarea-thatis2Kmawayfromthedanger-areinformedinafewtenthsofsecond,whentheythushadthechanceofmovingofafewmetersonly.VI.CONCLUSIONInthispaperweexploredthedesignspaceofbroadcastcommunicationservicesforVANETs,consideringtwokindsofalgorithms:theBEACONLESSone,thatreliessolelyoninstantaneousinformationonvehiclesposition,andtheBEA-CONEDone,thatmaintainsandexploitslonger-termknowl-edgeofthevehicularnetworktopology.Byextensivesimulation,wecomparedthedifferentap-proachesunderarealisticmicroscopictrafcmodel.Resultsshowthat,inthecaseofidealwirelesschannel,byprop-erlytuningthesystemparametersbothalgorithmsaccuratelypropagateabroadcastmessage.Moreover,ifthebeaconingoverheadisneglected,theBEACONEDapproachismoreefcientintermsofthechannelutilization.Conversely,theredundancyconnaturaltotheprobabilisticdecisionmakestheBEACONLESSalgorithmintrinsicallymorerobustwithrespecttoerrorsduetothewirelesschannel.Twointerestingconclusiveconsiderationscanbedrawn.First,itisnotpossibletoclearlystateawinnerofthecontest:indeed,differentapplicationsorservicesmayprefertouseonestrategyortheotherdependingontheirspecicrequire-ments.Forexample,bybeingmoreaccurateandrobust,theBEACONLESSapproachissuitableforroadsafetyapplica-tions.Second,andmostimportant,weshowedthathybridapproachesareapromisingalternative,astheyprovidethefreedomofnelytuningtheperformancetradeoffsbasedontheapplicationneeds.REFERENCES[1]DedicatedShortRangeCommunications(DSRC),http://www.leearmstrong.com/DSRC/DSRCHomeset.htm[2]eSafetyForum,http://europa.eu.int/information_society/programmes/esafety/text_en.htm[3]CarTALK2000,http://www.cartalk2000.net[4]Fleetnet,http://www.fleetnet.de[5]Q.Xu,T.Mak,J.Ko,R.Sengupta,“Vehicle-to-VehicleSafetyMessag-inginDSRC,”Proc.ofACMVANET'04,Page(s):19-29,2004[6]B.WilliamsandT.Camp,“ComparisonofBroadcastTechniquesforMobileAdHocNetworks,”'Proc.ofACMMOBIHOC'02,Jun.2002[7]Y.C.Tseng,S.Y.Ni,E.Y.Shih,“AdaptiveApproachestoRelievingBroadcastStormsinaWirelessMultihopMobileAdHocNetwork,”IEEETransactionsonComputers,May2003,Vol.52,No.5[8]S.Y.Ni,Y.C.Tseng,Y.S.Chen,J.P.Sheu,“Thebroadcaststormprobleminamobileadhocnetwork,”'Proc.ofACM/IEEEMOBICOM'99,1999.[9]G.Korkmaz,E.Ekici,F.Ozguner,U.Ozguner,“UrbanMulti-HopBroadcastProtocolforInter-VehicleCommunicationSystems,”Proc.ofACMVANET'04,Oct.2004.[10]E.Fasolo,R.Furiato,A.Zanella,“SmartBroadcastAlgorithmforInter-vehicularCommunications,”Proc.ofWPMC'05,Sep.2005.[11]H.Fuessler,H.Hartenstein,J.Widmer,M.Mauve,W.Effelsberg,Contention-BasedForwardingforStreetScenarios,Proc.ofWIT'04,Mar.2004.[12]M.Mauve,A.Widmer,H.Hartenstein,“Asurveyonposition-basedroutinginmobileadhocnetworks,”IEEENetworkMagazine,Vol.15,Iss.6,pp.30-39,Nov.-Dec.2001.[13]B.KarpandH.T.Kung,“GPSR:GreedyPerimeterStatelessRoutingforWirelessNetworks,”Proc.ofIEEE/ACMMOBICOM'00,Aug.2000.[14]S.Basagni,I.Chlamtac,V.R.Syrotiuk,B.A.Woodward,“Adistanceroutingeffectalgorithmformobility(DREAM),”Proc.ofIEEE/ACMMOBICOM'98,Oct.1998.[15]J.LeBrun,C.Chuah,D.Ghosal,M.Zhang,“Knowledge-BasedOppor-tunisticForwardinginVehicularWirelessAdHocNetworks,”Proc.ofIEEEVTC'05,Jun.2005.[16]H.AlshaerandE.Horlait,“AnOptimizedAdaptiveBroadcastSchemeforInter-vehicleCommunication,”Proc.ofIEEEVTC'05,Jun.2005.[17]R.Fracchia,M.MeoandD.Rossi,“OntheImpactofTrafcModelsonInter-vehicularBroadcastCommunications”Proc.ofIEEEMedHoc-Net'06,Jun.2006[18]S.PanwaiandH.Dia“ComparativeEvaluationofMicroscopicCar-FollowingBehavior”IEEETransactiononIntelligentTransportationSystems,Vol.6,No.3,Sep.2005[19]B.S.Kerner,“Experimentalfeaturesoftheemergenceofmovingjamsinfreetrafcow,”J.Phys.A33,,pp.221-228,2000.[20]P.G.Gipps,“Abehaviouralcarfollowingmodelforcomputersimula-tion,”TransportationRes.B,(15)1981,pp.105–111.[21]S.Krauß,P.Wagner.C.Gawron,“Metastablestatesinamicroscopicmodeloftrafcow,”,Phy.ReviewE,(55)1998[22]G.Anastasi,E.Borgia,M.Conti,E.GregoriandA.Passarella,“Un-derstandingtherealbehaviorofMoteand802.11adhocnetworks:anexperimentalapproach,”PervasiveandMobileComputing(Elsevier),vol.1,Jul.2005,pp.237–256.[23]M.ZunigaandB.Krishnamachari,“AnalyzingtheTransitionalRegioninLowPowerWirelessLinks,”Proc.ofIEEESECON'04,Oct.2004.[24]E.N.Gilbert,“Capacityofaburst-noisechannel,”BellSysr.Tech.J.,p.1253,Sept.1960.