基于GAIL方法的魚類個(gè)體運(yùn)動(dòng)策略恢復(fù)方法

打開文本圖片集

Individual fish movement strategy recovering approach based on GAIL

ONG Jinghan12，CHENPengyu，2，XUJun12，YUE Shengzhi12，MIN Zhongyuan2，LIU Xiaoyang12，LIYuanshan 2， 3

（1.School of Information Science and Engineering，Dalian Ocean University，Dalian 116O23，China; 2.KeyLaboratoryofMarineInformationTechnologyofLiaoningProvince，DalianOcean University，Dalian116O23，China; 3.KeyLaboratoryofEnvironmentControledquacultureMinistryofEducationDlianOceanUniversityalian6，Cina）

Abstract：Reinforcementlearninginfishbehaviorstrategiesfaces limitationssuchasbeingconstrainedbypredefinedrules， rewardfunctionsrelyingonpriorkowledge，andaninabilittofullcaptureobjectbehaviorstrategies.Inviewofthis，amethod basedongenerativeadversarialimitation learning（GAIL）isproposedtorecoverindividualmovementstrategiesbyfishswarm movementtrajectorydata.Thestateandactionrepresentationsof individualfisharedesigned，and thedecision-makingprocess offish movementisexpressedwithafullconnectedneuralnetwork.Experimentswereconductedwithonelearerandmultiple individualteacherswhonavigatewiththeVicsek model.ExperimentalresultsdemonstratethattheGAILmethodcanrecover individualfishmovementstrategiesefectively，providinganeffcientstrategylearningapproachappicabletothestudyand simulationofotherbiologicalswarmbehaviors.In-depthanalysisof theswarmbehaviorrevealstheinteractionrulesof individualsandgroupdyamics.Therefore，theproosedmethodofersnewinsightsfortheapplicationofartficialintellgencein biological behavior research.

Keywords：GAIL;fishscholbehavior;movement strategyrecovery;artificial intellgence aplication; Vicsek model; fully connected neural network

0 引言

科學(xué)領(lǐng)域的研究焦點(diǎn)[1-3]。（剩余8941字）