基于GAIL方法的鱼类个体运动策略恢复方法

打开文本图片集

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

科学领域的研究焦点[1-3]。（剩余8941字）