基于DE-MADDPG的多无人机协同追捕策略  

Cooperative pursuit strategy for multi-UAVs based on DE-MADDPG algorithm

作　　者：符小卫[1] 王辉 徐哲 FU Xiaowei;WANG Hui;XU Zhe(School of Electronics and Information,Northwestern Polytechnical University,Xi'an 710129,China)

机构地区：[1]西北工业大学电子信息学院,西安710129

出　　处：《航空学报》2022年第5期522-535,共14页Acta Aeronautica et Astronautica Sinica

基　　金：航空科学基金(202023053001)。

摘　　要：针对多无人机协同对抗快速目标的追逃博弈问题,研究了多无人机的协同追捕策略。基于解耦多智能体深度确定性策略梯度算法DE-MADDPG研究了多无人机协同对抗快速目标的追捕策略,设计了多无人机协同追捕的全局奖励和局部奖励两种奖励函数,训练后的多无人机能够有效地执行协同追捕任务。通过设置快速目标的多种逃逸控制策略,仿真验证了所设计的方法能够利用追捕无人机的数量优势,通过协作完成对快速目标的协同围捕,并且通过比较,验证本文所提出的算法相比MADDPG算法更快地取得了收敛效果。To solve the problem of pursuit-evasion game in multi-UAVs confronting the fast target,we study the cooperative pursuit strategy of multi-UAVs.We train the strategy using the DE composed Multi-Agent Deep Deterministic Policy Gradient(DE-MADDPG) algorithm,and design two reward functions:global reward function,and local reward function.The trained multi-UAVs can effectively carry out the cooperative pursuit mission.Simulation results show the effectiveness of the proposed method.The multi-UAVs can take advantage of numbers and cooperative work to complete a rounding up of the fast target.It is also verified that the proposed method can achieve faster convergence effect than the basic MADDPG algorithm.

关 键 词：多无人机 协同追捕 DE-MADDPG 多智能体强化学习 对抗策略 

分 类 号：V279[航空宇航科学与技术—飞行器设计]