Multi-UAV Cooperative Interception Planning in Obstacle Environments Based on Multi-Agent Reinforcement Learning
Jingrong Fang, Huayu Jia, Wenyue Meng, Bo Wang, Xiangyu Wang, J H Zhang
Chinese Academy of Sciences Institute of Engineering Thermophysics
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
To address the security challenge of intercepting Low-Slow-Small (LSS) aerial vehicles in dense urban building environments, where traditional interception methods are hindered by severe occlusion, this paper proposes a multi-UAV cooperative dynamic interception planning algorithm based on multi-agent reinforcement learning. The goal is to achieve reliable and safe interception of intruders through intelligent coordination of multiple high-speed fixed-wing UAVs. Building upon the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) framework, the proposed method introduces a composite state representation that explicitly incorporates repulsive perception from an artificial potential field, enabling agents to quantitatively sense the direction and magnitude of obstacle threats from local observations. A priority experience replay (PER) mechanism combined with cosine-annealing learning-rate scheduling is adopted to enhance policy convergence through prioritized sampling and dynamic learning-rate adjustment. Furthermore, a multi-objective reward function is designed to harmonize target approaching, obstacle avoidance, and formation maintenance. Simulation experiments demonstrate that the proposed algorithm achieves an interception success rate of$\mathbf{9 8. 0 8 \%}$with a zero collision rate over 1,200 randomly generated dense-obstacle test scenarios. The results validate the effectiveness, robustness, and safety of the framework for intelligent cooperative interception in complex urban settings, providing a technical reference for the development of intelligent UAV countermeasure systems in urban security.
逐年被引趋势
暂无年度引用数据
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
工程Guidance and Control Systems
Reinforcement Learning in Robotics · Robotic Path Planning Algorithms