智能城市交通系统外文翻译文献Word文档下载推荐.docx
《智能城市交通系统外文翻译文献Word文档下载推荐.docx》由会员分享,可在线阅读,更多相关《智能城市交通系统外文翻译文献Word文档下载推荐.docx(18页珍藏版)》请在冰点文库上搜索。
JohnFranceandAliA.Ghorbani
FacultyofComputerScience
UniversityofNewBrunswick
Fredericton,NB,E3B5A3,Canada
Abstract
Forthepurposesofmanaginganurbantrafficsystem,ahierarchicalmultiagentsystemthatconsistsofseverallocallyoperatingagentseachrepresentinganintersectionofatrafficsystemisproposed.LocalTrafficAgents(LTAs)areconcernedwiththeoptimalperformanceoftheirassigned
intersection;
however,theresultingtrafficlightpatternsmayresultinthefailureofthesystemwhenexaminedatagloballevel.Therefore,supervisionisrequiredandachievedwiththeuseofaCoordinatorTrafficAgent(CTA).ACTAprovidesameansbywhichtheoptimallocallightpatterncanbecomparedagainsttheglobalconcerns.Thepatterncanthenbeslightlymodifiedtoaccommodatetheglobalenvironment,whilemaintainingthelocalconcernsoftheintersection.
Functionalityoftheproposedsystemisexaminedusingtwotrafficscenarios:
trafficaccidentandmorningrushhour.Forbothscenarios,theproposedmultiagentsystemefficientlymanagedthegradualcongestionofthetraffic.
1Introduction
The20thcenturywitnessedtheworldwideadoptionoftheautomobileasaprimarymodeoftransportation.Coupledwithanexpandingpopulation,present-daytrafficnetworksareunabletoefficientlyhandlethedailymovementsoftrafficthroughurbanareas.Improvementstoroadnetworksareoftenconfinedbytheboundariesofexistingstructures.Therefore,theprimaryfocusshouldbetoimprovetrafficflowwithoutchangingthelayoutorstructureoftheexistingroadways.Anysolutiontotrafficproblemmusthandlethreebasiccriteria,including:
dynamicallychangingtrafficpatterns,occurrenceofunpredictableevents,andanon-finitebasedtrafficenvironment[2].Multiagentsystemsprovidepossiblesolutionstothisproblem,whilemeetingallnecessarycriteria.Agentsareexpectedtoworkwithinareal-time,non-terminatingenvironment.Aswell,agentscanhandledynamicallyoccurringeventsandmaypossesseveralprocessestorecognizeandhandleavarietyoftrafficpatterns
[3,5].
Althoughseveralapproachestodevelopingamultiagenttrafficsystemhavebeenstudied,eachstressestheimportanceoffindingabalancebetweenthedesiresofthelocaloptimumagainstamaintainedaverageatthegloballevel[4].Unfortunately,systemsdevelopedtoonlyexamineandoptimizelocaleventsdonotguaranteeaglobalbalance[6].However,localagentsarefullycapableofdeterminingtheirownlocaloptimum.Therefore,amorepowerfulapproachinvolvesthecreationofahierarchicalstructureinwhichahigher-levelagentmonitorsthelocalagents,andisabletomodifythelocaloptimumtobettersuittheglobalconcerns[7].
Theremainderofthispaperisorganizedasfollows.Section2examinestheproblemsofurbantraffic.ThedesignofahierarchicalmultiagentmodelisgiveninSection3.TheexperimentalresultsarepresentedinSection4.Finally,theconclusionsofthepresentstudyaresummarizedinSection5.
2UrbanTrafficCongestion
Improvementstourbantrafficcongestionmustfocusonreducinginternalbottleneckstothenetwork,ratherthanreplacingthenetworkitself.Ofprimaryconcernistheoptimizationofthetrafficlights,whichregulatethemovementoftrafficthroughthevariousintersectionswithintheenvironment.
Atpresent,trafficlightsmaypossesssensorstoprovidebasicinformationrelatingtotheirimmediateenvironment.Thisincludesroadandclocksensors,measuringthepresenceanddensityoftrafficandprovidingthetimeofdaytothetrafficlight.
Asolutiontotheurbantrafficproblemusingagentsistosimplyreplacealldecision-makingobjectswithinthesystembyacorrespondingagent.Eventhemostbasicsystemwillconsistofseveralagents,leadingtothecreationofamultiagentenvironment.Inthiscase,thetrafficenvironmentisbrokendownintoitsfundamentalcomponents,withoneagentforeachofthetrafficlightswithinthesystem.TomaintainorganizationandcooperationbetweentheLocalTrafficAgents(LTA),aCoordinatorTrafficAgent(CTA)existstomonitorglobalconcernsandmaintainorder.
3HierarchicalMultiagentModelforUrbanTraffic
Toachieveabalancebetweenthelocalandglobalaspectsofanurbantrafficsystem,amultiagentsystembasedonahierarchicalarchitectureisproposed.LTAsandCTAsmakeupthefundamentallevelsofthehierarchy,inwhichtheLTAsmeettheneedsofthespecificintersection,andtheCTAsdetermineifthechosenpatternsofaLTAaresuitedtomeetanyglobalconcerns.AsolitaryGlobalTrafficAgent(GTA)mayexistfornetworksofsufficientsize,andanInformationTrafficAgent(ITA)providesacentrallocationforthestorageofallsharedinformationwithinthesystem.Foreachagent,thevariablesnecessarytoorganizeandmaintainthehierarchyarelisted.
Thedevelopmentofthissystem,inwhichseveralLTAsworkundertheguidanceofasingleCTA,representsthebackbonetoahierarchicalstructureofagentswithinthesystem.TheCTAprovidesthebondsbetweenitselfandtheLTAsofthesystem,requiringthattheCTAstorealistoftheneighboringintersectionsforeachoftheLTAs.However,thecomputationalcapabilitiesofasingleCTAarelimited,andaroadnetworkofsufficientsizemayrequiretheuseofmultipleCTAstohandlealloftheLTAswithinthesystem.Inthiscircumstance,thenetworkwillbesubdividedintoregionscontrolledbyasingleCTA,withatop-levelGlobalTrafficAgent(GTA)linkingtheCTAstogether.TheGTAisanoptionalagent,existingonlyifthenetworkissufficientlylargethatitisrequired.
ALTAinteractsatagloballevelbysendingamessagecontainingthecalculatedoptimallocallightpatterntoitssupervisingCTA.TheCTAwillfindtheappropriateneighboringintersections,andthendeterminewhattheglobaloptimumforthehandledLTAwillbe.Tocalculatetheglobaloptimum,theCTAwillrequireallinformationrelatingtoeachoftheneighboringintersection.TheCTAwillrequesttheinformationfromtheITAbyprovidingalistoftheintersectionstheCTAisconcernedwith.Oncethisinformationisretrieved,aCTAcalculatestheglobaloptimumanddeterminesifavarianceexistsbetweenthelocalandglobaltrafficlightpatterns.Ifasignificantdifferenceisfound,abalancebetweenthelocalandglobaloptimumsmustbenegotiated,andthenreturnedtotheLTA.
4Implementation
TheproposedurbantrafficmultiagentsystemhasbeenimplementedusingtheJACKDevelopmentEnvironment,utilizingJACKIntelligentAgentsTM.JACKusestheBeliefDesireIntention(BDI)model.Underthisframework,“theagentpursuesitsgivengoals(desires),adoptingappropriateplans(intentions)accordingtoitscurrentsetofdata(beliefs)aboutthestateoftheworld.”[1].AgentscreatedundertheJACKenvironmentareevent-driven,andcanrespondtointernalorexternaleventsoccurringwithinthesystem
ThefirstphaseofimplementingthemultiagentsysteminvolvesthecreationofLTAs.Eachoftheseagentsaretailoredtomeettherequirementsofitscorrespondingintersection.Forthepurposesofthisproject,thetrafficnetworkconsistsofsixintersections.Eachintersectionconsistsoftworoadscrossingoveroneanother.Eachapproachingroadpossestwolanes,aleft-turninglane,andastraight/rightturninglane.
Thedecision-makingcapabilitiesoftheLTAsisdevelopedinthesecondphase.ThefirstroundofdecisionsbyaLTAareconcernedwithfindingthelocaloptimum,withnoconsiderationforneighboringintersections.Abasicexpertsystemdividesthesensorinputsintoacorrespondinglightpattern.Theresultinglightpatternconsistsofaneight-elementarray,whichcanbebrokendownintotwoelementsforeachoftheNorth,East,SouthandWestdirections.
Oddelementsofthearray(zeroisthefirstindex)specifythedurationoftheadvancedgreenstateforeachoftheappropriatedirections,whileevenelementsindicatethetimeofthestraight/right-turninglanes.Thislightpatternisalwaysinthesameformat,andoncecalculated,storedbytheLTA.Thevaluescontainedwithinthearrayconsistofstrings,indicatingthedurationofthetrafficlight.Thevaluesofthestringsareasfollows:
Red:
Redlight,lanesremaininastoppedstate.
Short:
Greenlight,mostfrequentlyoccurring,30-secondsindurationforstraightdirections,15secondsforleftturninglanes.
Medium:
Greenlight,oftenforaboveaveragetrafficdensities,45-secondsindurationforstraightdirections,25secondsforleft-turninglanes
Greenlight,indicatingahightrafficdensity,60-secondsindurationforstraightdirections,35secondsforleft-turninglanes.
Oncetheoptimallocaltrafficlightpatterniscalculated,theLTAsendsamessageeventtotheCTA.ThetrafficlightpatternispassedtotheCTA,allowingtheCTAtoadjusttheLTA’slightpatterntobettermeetanyglobalconcerns.StoredwithintheCTAisavectorofneighborsforeachLTAwithinthesystem.WhenaCTAreceivesamessageeventfromaLTA,theCTAgathersallinformationrelatingtotheneighborsofthecurrentlyhandledLTAfromtheITA.TheCTAwillusethisinformationwithinitsownexpertsystem,comparingthelocaloptimumlightpatternagainstthecurrentdensitiesoftheneighboringintersections.Ifasignificantdifferenceisfoundbetweenthelocaloptimumandtheessenceoftheglobaloptimum,thetrafficlightpatterntobeimplementedisalteredtoreducethedifferencebetweenthetwooptimums.ThenewtrafficlightpatternisreturnedtotheLTAforimplementationwithinthetrafficlight