张侃侃单片机控制交通灯参考文献.docx
《张侃侃单片机控制交通灯参考文献.docx》由会员分享,可在线阅读,更多相关《张侃侃单片机控制交通灯参考文献.docx(35页珍藏版)》请在冰豆网上搜索。
张侃侃单片机控制交通灯参考文献
原文出处
AustralasianTransportResearchForumAdelaide
FUZZYLOGICTRAFFICSIGNAL
CONTROL
ZEESHANRAZAABDY
PREPAREDFOR
DRNEDALT.RATROUT
INTRODUCTION
Signalcontrolisanecessarymeasuretomaintainthequalityandsafetyoftrafficcirculation.Furtherdevelopmentofpresentsignalcontrolhasgreatpotentialtoreducetraveltimes,vehicleandaccidentcosts,andvehicleemissions.Thedevelopmentofdetectionandcomputertechnologyhaschangedtrafficsignalcontrolfromfixed-timeopen-loopregulationtoadaptivefeedbackcontrol.Presentadaptivecontrolmethods,liketheBritishMOVA,SwedishSOS(isolatedsignals)andBritishSCOOT(area-widecontrol),usemathematicaloptimizationandsimulationtechniquestoadjustthesignaltimingtotheobservedfluctuationsoftrafficflowinrealtime.Theoptimizationisdonebychangingthegreentimeandcyclelengthsofthesignals.Inarea-widecontroltheoffsetsbetweenintersectionsarealsochanged.Severalmethodshavebeendevelopedfordeterminingtheoptimalcyclelengthandtheminimumdelayatanintersectionbut,basedonuncertaintyandrigidnatureoftrafficsignalcontrol,theglobaloptimumisnotpossibletofindout.
Asaresultofgrowingpublicawarenessoftheenvironmentalimpactofroadtrafficmanyauthoritiesarenowpursuingpoliciesto:
−managedemandandcongestion;
−influencemodeandroutechoice;
−improvepriorityforbuses,tramsandotherpublicservicevehicles;
−providebetterandsaferfacilitiesforpedestrians,cyclistsandothervulnerableroadusers;
−reducevehicleemissions,noiseandvisualintrusion;and
−improvesafetyforallroadusergroups.
Inadaptivetrafficsignalcontroltheincreaseinflexibilityincreasesthenumberofoverlappinggreenphasesinthecycle,thusmakingthemathematicaloptimizationverycomplicatedanddifficult.Forthatreason,theadaptivesignalcontrolinmostcasesisnotbasedonpreciseoptimizationbutonthegreenextensionprinciple.Inpractice,uniformityistheprinciplefollowedinsignalcontrolfortrafficsafetyreasons.Thissetslimitationstothecycletimeandphasearrangements.Hence,trafficsignalcontrolinpracticearebasedontailor-madesolutionsandadjustmentsmadebythetrafficplanners.Themodernprogrammablesignalcontrollerswithagreatnumberofadjustableparametersarewellsuitedtothisprocess.Forgoodresults,anexperiencedplannerandfine-tuninginthefieldisneeded.Fuzzycontrolhasproventobesuccessfulinproblemswhereexactmathematicalmodellingishardorimpossiblebutanexperiencedhumancancontroltheprocessoperator.Thus,trafficsignalcontrolinparticularisasuitabletaskforfuzzycontrol.Indeed,oneoftheoldestexamplesofthepotentialsoffuzzycontrolisasimulationoftrafficsignalcontrolinaninter-sectionoftwoone-waystreets.Eveninthisverysimplecasethefuzzycontrolwasatleastasgoodasthetraditionaladaptivecontrol.Ingeneral,fuzzycontrolisfoundtobesuperiorincomplexproblemswithmultiobjectivedecisions.Intrafficsignalcontrolseveraltrafficflowscompetefromthesametimeandspace,anddifferentprioritiesareoftensettodifferenttrafficflowsorvehiclegroups.Inaddition,theoptimizationincludesseveralsimultaneouscriteria,liketheaverageandmaximumvehicleandpedestriandelays,maximumqueuelengthsandpercentageofstoppedvehicles.So,itisverylikelythatfuzzycontrolisverycompetitiveincomplicatedrealintersectionswheretheuseoftraditionaloptimizationmethodsisproblematic.
Benefitsanddisadvantagesoffuzzysystems
Fuzzylogichasbeenintroducedandsuccessfullyappliedtoawiderangeofautomaticcontroltasks.Themainbenefitoffuzzylogicistheopportunitytomodeltheambiguityandtheuncertaintyofdecision-making.Moreover,fuzzylogichastheabilitytocomprehendlinguisticinstructionsandtogeneratecontrolstrategiesbasedonprioricommunication.Thepointinutilizingfuzzylogicincontroltheoryistomodelcontrolbasedonhumanexpertknowledge,ratherthantomodeltheprocessitself.Indeed,fuzzycontrolhasproventobesuccessfulinproblemswhereexactmathematicalmodellingishardorimpossiblebutanexperiencedhumanoperatorcancontrolprocess.Ingeneral,fuzzycontrolisfoundtobesuperiorincomplexproblemswithmulti-objectivedecisions.
Atpresent,thereisamultitudeofinferencesystemsbasedonfuzzytechnique.Mostofthem,however,sufferill-definedfoundations;eveniftheyaremostlyperformingbetterthatclassicalmathematicalmethod,theystillcontainblackboxes,e.g.defuzzification,whichareverydifficulttojustifymathematicallyorlogically.Forexample,fuzzyIF-THENrules,whichareinthecoreoffuzzyinferencesystems,areoftenreportedtobegeneralizationsofclassicalModusPonensruleofinference,butliterallythisnotthecase;therelationbetweentheserulesandanyknownmany-valuedlogiciscomplicatedandartificial.Moreover,theperformanceofanexpertsystemshouldbeequivalenttothatofhumanexpert:
itshouldgivethesameresultsthattheexpertgives,butwarnwhenthecontrolsituationissovaguethatanexpertisnotsureabouttherightaction.Theexistingfuzzyexpertsystemsveryseldomfulfilthislattercondition.
Manyresearchesobserve,however,thatfuzzyinferenceisbasedonsimilarity.Kosko,forexample,writes'Fuzzymembership...representssimilaritiesofobjectstoimpreciselydefinedproperties'.Takingthisremarkseriously,westudysystematicallymany-valuedequivalence,i.e.fuzzysimilarity.Itturnsoutthat,startingfromtheLukasiewiczwell-definedmany-valuedlogic,weareabletoconstructamethodperformingfuzzyreasoningsuchthattheinferencereliesonlyonexpertsknowledgeandonwell-definedlogicalconcepts.Thereforewedonotneedanyartificialdefuzzificationmethod(likeCenterofGravity)todeterminethefinaloutputoftheinference.Ourbasicobservationisthatanyfuzzysetgeneratesafuzzysimilarity,andthatthesesimilaritiescanbecombinedtoafuzzyrelationwhichturnsouttoafuzzysimilarity,too.Wecallthisinducedfuzzyrelationtotalfuzzysimilarity.FuzzyIF-THENinferencesystemsare,infact,problemsofchoice:
compareeachIF-partoftherulebasewithanactualinputvalue,findthemostsimilarcaseandfirethecorrespondingTHEN-part;ifitisnotunique,useacriteriagivenbyanexperttoproceed.BasedontheLukasiewiczwelldefinedmanyvaluedlogic,weshowhowthismethodcanbecarriedoutformally.
HypothesisandPrinciplesofFuzzyTrafficSignalControlTrafficsignalcontrolisusedtomaximizetheefficiencyoftheexistingtrafficsystems[6].However,theefficiencyoftrafficsystemcanevenbefuzzy.Byprovidingtemporalseparationofrightsofwaytoapproachingflows,trafficsignalsexertaprofoundinfluenceontheefficiencyoftrafficflow.Theycanoperatetotheadvantageordisadvantageofthevehiclesorpedestrians;dependonhowtherightsofwaysareallocated.Consequently,theproperapplication,design,installation,operation,andmaintenanceoftrafficsignalsiscriticaltotheorderlysafeandefficientmovementoftrafficatintersections.
Intrafficsignalcontrol,wecanfindsomekindofuncertaintiesinmanylevels.Theinputsoftrafficsignalcontrolareinaccurate,andthatmeansthatwecannothandlethetrafficofapproachesexactly.Thecontrolpossibilitiesarecomplicated,andhandlingthesepossibilitiesareanextremelycomplextask.Maximizingsafety,minimizingenvironmentalaspectsandminimizingdelaysaresomeoftheobjectivesofcontrol,butitisdifficulttohandlethemtogetherinthetraditionaltrafficsignalcontrol.Thecauseconsequence-relationshipisalsonotpossibletoexplainintrafficsignalcontrol.Thesearetypicalfeaturesoffuzzycontrol.
Fuzzylogicbasedcontrollersaredesignedtocapturethekeyfactorsforcontrollingaprocesswithoutrequiringmanydetailedmathematicalformulas.Duetothisfact,theyhavemanyadvantagesinrealtimeapplications.Thecontrollershaveasimplecomputationalstructure,sincetheydonotrequiremanynumericalcalculations.TheIFTHENlogicoftheirinferencerulesdoesnotrequiremuchcomputationaltime.Also,thecontrollerscanoperateonalargerangeofinputs,sincedifferentsetsofcontrolrulescanbeappliedtothem.IfthesystemrelatedknowledgeisrepresentedbysimplefuzzyIFTHEN-rules,afuzzy-basedcontrollercancontrolthesystemwithefficiencyandease.Themaingoaloftrafficsignalcontrolistoensuresafetyatsignalizedintersectionsbykeepingconflicttrafficflowsapart.Theoptimalperformanceofthesignalizedintersectionsisthecombinationoftimevalue,environmentaleffectsandtrafficsafety.Ourgoalistheoptimalsystem,butweneedtodecidewhatattributesandweightswillbeusedtojudgeoptimality.
Theentireknowledgeofthesystemdesignerabouttheprocess,trafficsignalcontrolinthiscase,tobecontrolledisstoredasrulesintheknowledgebase.Thustheruleshaveabasicinfluenceontheclosed-loopbehaviourofthesystemandshouldthereforebeacquiredthoroughly.Thedevelopmentofrulesistimeconsuming,anddesignersoftenhavetotranslateprocessknowledgeintoappropriaterules.SugenoandNishidamentionedfourwaystoderivefuzzycontrolrules:
1.operatorsexperience
2.controlengineer'sknowledge
3.fuzzymodellingoftheoperator'scontrolactions
4.fuzzymodellingoftheprocess
Zimmermannaddedthreesourcesmore
5.crispmodelingoftheprocess
6.heuristicdesignrules
7.on-lineadaptationoftherules.
Usuallyacombinationofsomeofthesemethodsisnecessarytoobtaingoodresults.As