局部信息下的分布式无人集群自组织编队框架

打开文本图片集

关键词：分布式无人集群；局部邻域信息；复杂网络；多智能体强化学习；去中心化策略优化；自组织编队中图分类号：TP311.5文献标志码：A 文章编号：1001-3695（2026）01-026-0216-11doi：10.19734/j. issn.1001-3695.2025.06.0175

Distributed unmanned swarm self-organizing formation framework based on local information

Cui Zihan，Liu Wei†，XieWanzhen，Hu Diwei （SchoolofComputer Science&Engineering，Wuhan InstituteofTechnology，Wuhan43O2O5，China）

Abstract：This paper proposedadual-layer adaptiveformationframework toadresslowlocal information transmissionefficiency，slow individualresponse，andpoorglobalcoordinationindistributedunmannedswarmsunderdynamicandcomplex environmentswithcommunicationconstraints.Theframeworkbuiltadistributednetwork modelcombiningacommunication layerandastructurallayer.Inthecommunicationlayer，itappliedmulti-agentreinforcementlearningwithdecentralizedpolicyoptimization basedonlocalneighborhood informationtoachieveefficient information sharingand policyupdating.Inthe structurallayer，itintroducedadistributedtopologyreconfigurationmechanismtoenableflexibleadjustmentandeconstruction offormations indamagerecoveryandtask-splitingscenarios.Itembeddedadynamicdisturbancehandlingmechanismbetween thetwolayers toachieverapidadaptationandreconfigurationfornodefailuresandtaskchanges.Simulationresultsshowthat the proposed methodsignificantlyimproves task successratesandconvergence speed inanti-attck recoveryand formationspliting tasksacrossmultiple network topologies.Theframework demonstrates strong robustnessandadaptability，providingan efectiveapproachandtheoreticalsupportforeffcientcoaborationandstablefomationofunmannedswarmsunderlocalinformation constraints.

Keywords：distributedunmannedswarm；localneighborhoodinformation；complexnetworks；multi-agentreinforcement learning；decentralized strategy optimization； self-organizing formation

0 引言

无人集群系统（unmanned swarm system，USS）是由大量同构或异构智能体（如无人机、无人车、无人船等）及其配套的传感器、计算单元、通信设备构成的智能化系统。（剩余25340字）