警用无人机协同巡逻的多模态ALNS优化算法

打开文本图片集

中图分类号：TP18；V279 文献标识码：A 文章编号：1006-8228（2025）12-56-07

Abstract：ThispaperaddresesthetypicalNP-hardproblemofcooperativepatrolpathplaningforpoliceunmanedaerialvehicles （UAVs）byproposinganovelchaotic-reinforcement-leaing-drivenmultimodaladaptivelargeneighborhoodsearch（CRL-ALN） algorithm.Thealgorithmincorporatesadeepreinforcementlearingframeworktoestablishanintellgentmultimodaloperator selectionmechanism，whichdynamicallcordinatesdestructionandrepairoperatorswhileintegratingchaoticsequencestoenhance globalexplorationcapabilityFormuli-AVcooperativepatrolscenariosfromasinglenest，adualobjectiveoptimizationmodelis formulatedwithcomprehensiveconsiderationoftimewindowconstraintsandpriort-basedcoverage.Extensiveexperimentsonten randomlygeneratedtestinstancesdemonstratethattheproposedCRL-ALNSalgorithmachievessignificantperformance improvements over an enhanced genetic algorithm，reducing total flight distance by 72.7% and the number of required UAVs by 38.5% onaveragewhilemaintainingstabilityandconvergenceacrossdiversespatialconfigurations，thusfullyvalidatingits effectiveness and superiority.

KeyWords：PoliceUAV；PathPlanning;ALNS；ReinforcementLearning;ChaoticOptimization;Multimodal OperatorSelectic

0引言

当前公共安全形势日益复杂，构建立体化治安巡逻体系具有重要现实意义。（剩余8345字）