基于多機(jī)器人系統(tǒng)的避碰規(guī)劃再勵(lì)學(xué)習(xí)研究

1、揚(yáng)州大學(xué)碩士學(xué)位論文基于多機(jī)器人系統(tǒng)的避碰規(guī)劃再勵(lì)學(xué)習(xí)研究姓名：唐英申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)技術(shù)指導(dǎo)教師：楊月全20091001揚(yáng)州大學(xué)碩士學(xué)位論文2AbstractWiththedevelopmentoftllemultirobotsystemtechnologytheabilityofrobotscontinuouslyisimprovedanditsapplicationfieldsareexpandedWhen柚envir

2、onmentofcooperativemobilerobotsbc；c‘蛐窿sinocecomplex，muhiplemobilerobotssystclllswhichenmposedofmanyrobotsareusedtocompletetheworkwhichasinglerobotcannotcompleteThemobilerobotsmayworkintheunknov／nandcomplexenvironmentRobo

3、tmotioncontroltmghonreinforcementlearningneednotpreciseinformationofrobotmovements，butdestinationlocationofrobotsinthetaskReinforcomemIe盯ningis丑kindofreal—timeandonlinelearning丑Ig‘蛹lllmThroughmachinelearning，robotsCallga

4、inuseful麟耐a鸛，andmas燈柚optimalstrategyandknowledgestepbystepThestudyinthisthesishasstrongapplicationdemandandapplicationvaluetesearchofcooperationbetweenmultirobotsForreinforuememlearningdt嘣atyil啦目疆貯姆withenvimnmcmandneedno

5、tenvironmentmodelanda咖knowledge，itisemployedineoope咖mobilerobotssystemThercinforcmentfunctionisdesignedwithbehaviorbasedandmodel—freeThemovablerobotsinthemobilerobotsystemavoidcollisionwiththedynamicandstaticobstaclesThe

6、maincontributioninthisthesisisgivenasfollowsFirsaythe枷tiesresearchstatusanddcvclopammtmmtmcfiveofmulti—robotsystems躺introducedsystematicallySecondlyamodifiedartificialfieldalgorithmbasedOilrobotspccdmarginalfunctionre(u)

7、ispresentedwhichisappliedincoo刪vemobilerobo伍systemsTimminfmrcmcntfunctionisverifiedinthesimulmiOntestsThirdlyallimprovedQ—learningalgorithmwithvariableattenuationfactorisproposedwhichcombinesadvantagesofthe8Greedystrateg