強化學習與決策樹方法在RoboCup中的研究與應(yīng)用.pdf

1、Y890917分類號；TP242題單位代碼學號密級北京工業(yè)大學碩士學位論文i0005S20030209l英文并列△里￡墜!g△!!Q塑Q￡必2匹Q8堡壘堡盟!墜星△B盥盟Q題目塵塵巴衛(wèi)基￡l璺!Q塑三墅星!邀基Q娶Q￡望￡研究生姓名：堂垂廷專業(yè)：蕉劍堡鎏望絲劍王捏研究方向：蟹能撞劍理監(jiān)量廑圍導(dǎo)師姓名：整堂畦職稱：副熬援論文報告提空目期2鯉生5丑學位授予只期授予單位名稱和地址jE塞王些丕蘭jk塞正塑旦墮巫圈!螋蘭北京工業(yè)大學工學碩士學位論

2、文AbstractTheresearchonMulti—Agentsystem(MAS)aIlddis砸butedamficialintelligencehasbecomemefocusofcomputcrscienceandartificialintelligenceTheMASconsistsofmanyagentsnmainly‰dsoutmesolutionfor也ereal一血neprocessofhlfomlation趾dh

3、amlonypmblemsundercompIcxanddyll鋤ice11viromentsRobocup(111eRobotworldCup)si舢1ationg鋤esyst鋤isagoods鋤pleoft11eresearchofanificialIntdligcncea11drobot，anditisagoodrcsearchplatfomfort11eresearch0fMASBecausesomeimpo恤ntcharact

4、e璐suchasdyllamic，realtime，competi廿ve，not如lliIlfo衄ation髓dsoon，agents’leanlillgaIldmulti—agentcoordinationhavebecomememaindircc廿onofrcsearchers111is也csisfo∞sesont11edesi驢aIldimpl鋤entofMultiAgerltSystem岫RoboCup舳crprese毗培血ec

5、onc印tofM111ti—AgentSystemaIldmebackgml】ndofRoboCup，meaimofthismesis，howtofbllIldaRoboCupS婦ulationCompedtionte鋤namedBJUTisin仃o(1ucedindetailThemaincontIibutionsofmist11esisinclude：First，theclassicalagents鋤cllitecnIreisuse

6、dinBJUT11lisarchitectureconsigtsofmodeling，communication，Perc印tionIntc印rctation，deliberativere船on協(xié)gaIldreactivemoduleandsoon111tllisway協(xié)earchitcctI】rec姐decreasedi珩cul廿esofdesigning血ewholeagentSecond，mesysteIIlarcllitecn鵬

7、ofBJUTispresentedwhichcanbeusedinreal—time，noisy，advers撕a1a11dpressiIlgmmti—agents”tcnl砌rd，thededsion仃eealgo珊Hnisin拄DducedindetailApplyanewC5Oalgorit№basedonittos01vepassproblemiIlRoboCupAgenttakesactionofpaSs麗tllusingde

8、dsiontreepmducedbymisa190riⅡlm，andadjustscon仃0lfactorsbytakingomeractionsFinalIyitimpmvesmesuccessmtioofpassFounh，Q—LeamiIlgalgod【llmisirⅣestigatedtes01vemedribblmgballpmbl鋤inRoboCup111ismemodchallgesnleactionselectionof

9、traditionalg嘶mm，a11d僦nsa11酎eandpowerofactioni11steadofactionsdectionThetcstsproVeitsavajlabili吼pLtlast，rcinforcemellt1e鋤ing印proachbasedonASPLmodeltocoope枷Vebcha_vioracquisitioniⅡmulti—agentsyst咖is印pliedtos01Vetllepmbl鋤of