基于MULTIAGENT的城市交通信号控制研究

《基于MULTIAGENT的城市交通信号控制研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、长沙理工大学硕士学位论文基于Multi-Agent的城市交通信号控制研究姓名：刘虹秀申请学位级别：硕士专业：交通信息工程及控制指导教师：魏武20080326ABSTRACTWiththecontinuousdevelopmentofnationaleconomy,people’Slivingstandardimprovesgradually,vehiclepopulationgrowsunceasingly,andurbantrafficproblembecomesobviouslydaybyday．Thetrafficsystemisacomplicatedla

2、rgescalesystemwithrandomicityandindefinability,Thecurrentsignalcontrolapproachhasobviousshortcomingsinreliabilityofcontrolandglobaloptimizingandwillunabletosolvetheincreasingtrafficproblem，makeuseofmoderntechnologyandintelligentmethodtosolveurbantrafficproblemhasattractedmuchattention

3、ofresearchers．Withthedevelopmentofdistributedartificialintelligencetechnology，AgenttechnologyandMulti—agentsystemtheoryhasprovidedatheoryfoundationforthestudyofsoftwareintelligentunderdistributedenvironment，astheurbantrafficsignalcontrolisinherentlydistributed，ithasverygoodprospecttoa

4、doptthemulti-Agenttechnologytostudyurbantrafficsignalcontrolproblem．Inthispaper,combinedwithAgenttechnology，adistributedandcoordinatedtrafficsignalcontrolsystembasedonUTSCAisproposed．Firstly,thispapercarriedonstructuraldesignforUTSCAatsingleintersection，describeditsworkprocess，andhasc

5、arriedonthedesigntothestudypatternofthestudyunit，andthen，carriedonthestructuraldesignforthetrafficsignalcontrolsystembasedonUTSCA，andtheformaldescriptionofthesystem．Then，selectstrafficstatesofsingleintersection，formsquantitativedescriptionforthearrivedcarsandsignalstatesofintersection

6、throughfuzzyclassifying，andconstructsthesignalcontrolrules，takesthetotalstopdelayofcarsfortheoptimizedgoal，usestheimprovedQ-learningalgorithmtocarryontheAgenttrainingtoimprovethesignalcontrolrule．Bycomparingthemethodwithtraditionalmethodthroughsimulation．simulationresultindicatesthatt

7、heeffectofthenewmethodinisolatedintersectioniSbetter．Finally．carriedonthedescriptionofthecommunication，coordinatedway,andlearningmodelinthesignalcontrolsystembasedonMulti—agent，usesthedistributedreinforcementlearningmethodforcoordination，thesimulationresultindicatesthatthemethodcanred

8、uceth