复杂环境下无人机集群的协同控制研究

打开文本图片集

中图分类号：TP181；V279 文献标识码：A 文章编号：2096-4706（2025）19-0043-07

Abstract： Facing the challenges of complex electromagnetic interference，dynamic obstacles and communication security，thispaperproposesahybridframeworkofBionic Hierarchical-Multi-AgentReiforcementLearning（BioRL-Swarm）. Through thecolaborative control model of dynamic Hierarchical Network Coupling （HNC）and Multi-Agent Proximal Policy Optimization （MAPPO），itrealizes the dp integrationofdistributed decision-making mechanismandsurvivability communication，andtheclusterautonomyandenvironmentaladaptabilityaresignifcantlyimproved.Simulationresultsshow thatcomparedwiththe traditionalalgorithm，BioRL-Swarmimproves thesuccessrateofdynamicobstacleavoidanceby 23.6% reduces the communication packet loss rate to 1.2% ，and shortens the task execution time by 37.4% ，which showssignificant advantages inmanykey indicators.It provides asolid theoretical basisandreliable technicalverificationforthe engieering application of intelligent unmanned systems in military， agricultural and other fields.

Keywords： UAVcluster; cooperative control; DeepReinforcementLearning;bionic hierarchical network

0 引言

通过对现有文献的系统分析，发现当前无人机集群在复杂环境中面临着三大挑战：一是动态障碍物与电磁干扰的联合建模困难，导致避障成功率普遍低于70% ；二是传统通信协议（如时间分配多路访问协议（TDMA））在出现链路中断时恢复耗时超过 5s ，难以满足应急任务需求；三是集中式决策架构存在单点故障风险，当集群规模超过20架时，失控率将升至40% 。（剩余6238字）