多无人机协同路径规划计算资源分配方法研究  

Research on Resource Allocation Method for Multi-UAV Cooperative Path Planning

作　　者：聂铭涛 刘颖珂 程海峰 刘小雄[1] NIE Mingtao;LIU Yingke;CHENG Haifeng;LIU Xiaoxiong(School of Automation,Northwestern Polytechnical University,Xi'an 710072,China;Unit 94521,PLA,Zibo 255399,China)

机构地区：[1]西北工业大学自动化学院,西安710072 [2]中国人民解放军94521部队,山东淄博255399

出　　处：《计算机测量与控制》2025年第4期147-154,161,共9页Computer Measurement &Control

摘　　要：针对多目标多机协同路径规划问题,在MOEA/D算法求解的基础上,采用深度强化学习方法对MOEA/D算法中计算资源分配方法进行了研究;对多机协同路径规划问题进行了研究,分析了相关约束以及优化目标,建立多机协同路径规划多目标优化模型;结合协同进化思想,对基于分解的多目标进化协同路径规划进行了研究;对基于强化学习的计算资源分配策略进行了研究,实现了深度强化学习在多目标优化计算资源分配问题上的应用;实现了多机协同路径规划仿真验证;经仿真测试,算法以更高性能完成多机协同路径规划任务,提高了多机协同路径规划中计算资源分配策略的能力。Aiming at the problem of multi-objective and multi-UAV cooperative path planning,based on the solution of MOEA/D algorithm,a deep reinforcement learning method is adopted to study the computational resource allocation method in the MOEA/D algorithm;Study the multi-UAV cooperative path planning,analyze the relevant constraints and optimization objectives,and establish the multi-objective optimization model of multi-UAV cooperative path planning;Investigate the multi-objective evolutionary cooperative path planning based on decomposition by combined with the idea of co-evolution,study the computational resource allocation strategy based on reinforcement learning,realize the application of deep reinforcement learning in the multi-objective optimization of computational resource allocation,and achieve the simulation verification of the multi-UAV cooperative path planning;Though simulation test,the algorithm completes the multi-UAV cooperative path planning task with a higher performance and improves the performance of the computational resource allocation strategy.

关 键 词：多目标进化 深度强化学习 计算资源分配 协同路径规划 基于分解 

分 类 号：TP4081[自动化与计算机技术] V219[航空宇航科学与技术—航空宇航推进理论与工程]