基于MATD3的无人机智能协同路径规划方法  

Intelligent collaborative path planning method for unmanned aerial vehicles based on MATD3

作　　者：杨闰麟 陈才轶 郭正玉[2] 罗德林 YANG Runlin;CHEN Caiyi;GUO Zhengyu;LUO Delin(School of Aerospace Engineering,Xiamen University,Xiamen 361102,China;China Airborne Missile Academy,Luoyang 471000,China;National Key Laboratory of Air-based Information Perception and Fusion,Luoyang 471000,China)

机构地区：[1]厦门大学航空航天学院,福建厦门361102 [2]中国空空导弹研究院,河南洛阳471000 [3]空基信息感知与融合全国重点实验室,河南洛阳471000

出　　处：《厦门大学学报(自然科学版)》2025年第2期241-252,共12页Journal of Xiamen University:Natural Science

基　　金：空基信息感知与融合全国重点实验室与航空科学基金联合资助项目(20220001068001)。

摘　　要：[目的]多无人机协同路径规划是多无人机应用的关键技术之一,本文旨在研究三维动态环境下的多无人机智能协同路径规划和优化方法.[方法]本文提出了一种基于搜索的扰动流体动态系统(SIFDS)与速度控制策略结合的路径规划算法.该方法通过SIFDS方法将多障碍物环境简化为单障碍物环境,并生成由反应系数构成的无人机动作空间;再结合速度控制策略实现无人机间的时间协同;最后利用多智能体双延迟深度确定性策略梯度(multi-agent twin delaged deep deterministic policy gradient,MATD3)算法,训练得到的动作网络在线生成最优的反应系数,以此提高多无人机路径规划的质量与效率.[结果]本文所提出的方法能够在三维动态环境下使多无人机规划出安全且互不干扰的路径;引入MATD3后,无人机路径更短且平滑;速度控制策略能够控制各无人机同时到达的时间误差在测量精度0.5 s以下.[结论]本文所提出的方法,能够使多无人机在规划路径的同时满足时间协同、空间协同的约束,实现了多无人机协同路径规划,且具有较强的稳定性,为多无人机系统在复杂任务中的实际应用提供了一定的技术支持和理论参考.[Objective]Coordinated path planning for multiple unmanned aerial vehicles(multi-UAVs)represents one of the pivotal technologies in multi-UAV applications.In this study,we aim to investigate intelligent coordinated path planning and optimization methods for multi-UAVs in three-dimensional dynamic environments.[Methods]Herein we propose a path planning algorithm that combines search-based interfered fluid dynamic system(SIFDS)with a velocity control strategy.This method utilizes SIFDS to simplify a multi-obstacle environment into a single obstacle environment and generates a UAV action space composed of reaction coefficients.Also,the velocity control strategy is employed to achieve temporal coordination among UAVs.Combined with the multi-agent twin delayed deep deterministic policy gradient(MATD3)algorithm,the trained action network generates the optimal reaction coefficient online,thus improving the quality and the efficiency of multi-UAV path planning.[Results]The proposed method enables multi-UAVs to plan safe and non-interfering paths in a three-dimensional dynamic environment.After introducing MATD3,the path of the drone is shorter and smoother.Furthermore,comparative experiments illustrate that the integrated velocity control strategy facilitates the simultaneous arrival of all UAVs at their destinations.The arrival time error is below 0.5 seconds of measurement accuracy.[Conclusions]The proposed method allows multi-UAVs to meet both temporal and spatial collaboration constraints while planning their paths,so that effective collaborative path planning is achieved.Hopefully,our study can provide technical support and theoretical guidance for the practical deployment of multi-UAV systems in complex tasks.

关 键 词：无人机 协同路径规划 扰动流体动态系统 强化学习 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]