基于價(jià)值的機(jī)器學(xué)習(xí)方法及其在RoboCup仿真2D中的應(yīng)用.pdf

1、合肥工業(yè)大學(xué)碩士學(xué)位論文基于價(jià)值的機(jī)器學(xué)習(xí)方法及其在RoboCup仿真2D中的應(yīng)用姓名：李龍申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：王浩20090301LearningBasedonAxiologyanditsApplicationinRoboCup2DSimulationABSTRACTMachineLearningisaprojectforstudyandapplicationinArtificialIntelligence

2、，Robotics，NeuralNetwork，andSOonIthasgoodapplyingprospectinindus％medicine，militaryaffairs，sportsactivities，whichhavebecomecrucialtechnologyTherearemanymethodsaboutMachineLearningThefamiliararithmeticsincludereinforcementl

3、earning，NeuralNetworklearning，BayesLearning，anddecisiontreelearningThemostimportantofMachineLearningishowtotrainingadaptiveresultfastandeffectivelyConsideringthispoint，alearningmethodbasedonmaxvalueisintroducedinthisthes

4、is，andtheapplicationinRoboCup2DshowsitsgoodeffectivenessThemainworkislistedasfollowing：1Imitatingpeople’Sthinking，anewlearningmethodbasedonmaxvalueLBAisintroducedinthisthesisFirstlyworldmodelisdividedintoseveralfactorsan

5、dthenamendthevalueaccordingthesefactorsandoneactionFinallychoosethemaxvalueandmakedecision2Amethodtopredictagentpositionisdescribed，thatisEIA(TargetTraceAlgorithmbasedonExcludingIrrespectiveArea)EIAisanideabasedonprobabi

6、lityofactionItdefinesAgentConfidenceandAreaConfidenceanddecidesmoreexactpositionofagentbasedonthedifferencebetweenthesetwoconfidences3ThoseabovemethodsisappliedinRoboCup2DteamdesignanddeviseavisiondecidemodelManyexperime