航空科学技术2026，Vol.37Issue(1)：14-23,10.DOI:10.19452/j.issn1007-5453.2026.01.002
基于双中心化Q网络框架的多无人机路径规划算法
Multi-UAV Path Planning Algorithm Based on Twin Decentralized Q-Network Framework
摘要
Abstract
Path planning,as the core component enabling autonomous collaborative operation in multi-UAV systems,directly determines the execution efficiency and safety of cluster missions.This paper investigates the homogeneous multi-UAV path planning problem and proposes a multi-agent reinforcement learning algorithm based on a dual-centralized Q-network framework to address this challenge.Initially,through modeling UAVs and obstacles.This paper analyzes collision risks,motion continuity constraints,and mission execution requirements during UAV flight,formulating the multi-UAV path planning problem as a multi-constrained combinatorial optimization problem with high computational complexity.Subsequently,inspired by the twin delayed deep deterministic policy gradient algorithm,this paper develops a multi-agent reinforcement learning approach employing a dual-centralized Q-network framework to automatically generate feasible and collision-free flight paths for each UAV.Simulation experiments validate the algorithm's effectiveness in multi-UAV path planning,providing an efficient solution for autonomous path planning in UAV swarms.关键词
双中心化Q网络框架/自适应贪婪策略/多无人机/路径规划Key words
twin centralized Q-network framework/adaptive greedy strategy/multi-UAV/path planning分类
航空航天引用本文复制引用
任崇德,陈进朝,赵爽,刘九如..基于双中心化Q网络框架的多无人机路径规划算法[J].航空科学技术,2026,37(1):14-23,10.基金项目
国家自然科学基金(62106202) （62106202）
航空科学基金(2023M073053003) （2023M073053003）
陕西省重点研发计划(2024GX-YBXM-118) National Natural Science Foundation of China(62106202) （2024GX-YBXM-118）
Aeronautical Science Foundation of China(2023M073053003) （2023M073053003）
Key Research and Development Program of Shaanxi Province(2024GX-YBXM-118) （2024GX-YBXM-118）
