ATT-CEL-MADDPG算法在多无人机围捕中的协同优化
成星宇

郑州航空工业管理学院，河南郑州，450046；

摘要：针对多无人机协同围捕任务的性能优化问题，文章提出 ATT-CEL-MADDPG 算法构建含动态障碍与部分可观测特性的二维仿真环境，据此评估各算法性能。实验表明，ATT-CEL-MADDPG 的注意力机制可有效建模智能体间的空间结构与协作关系，其收敛速度、最终任务成功率及环境适应性均显著优于对比方法；结合课程学习机制，能逐步引导智能体完成从追踪、包围到收网的协同围捕流程。结果验证了课程式经验学习与注意力增强 Critic 在提升多无人机协同围捕任务学习效率与策略稳定性中的有效性，为复杂动态环境下的多无人机协同决策提供了新的算法思路与实验依据。

关键词：无人机围捕；多智能体深度强化学习；ATT-CEL-MADDPG；Curriculum Learning

参考文献

[1]ATIF M, AHMAD R, AHMAD W, et al. UAV-assisted wireless localization for search and rescue[J]. IEEE Systems Journal, 2021, 15(3): 3261-3272.

[2]CHUANG H M, HE D Q, NAMIKI A. Autonomous target tracking of UAV using high-speed visual feedback[J]. Applied Sciences-Basel, 2019, 9(21): 4473.

[3]SUDHAKAR S, VIJAYAKUMAR V, KUMAR C S, et al. Unmanned aerial vehicle (UAV) based forest fire detection and monitoring for reducing false alarms in forest-fires[J]. Computer Communications, 2020, 149: 1-16.

[4]KUANG Q, JIN X, ZHAO Q P, et al. Deep multimodality learning for UAV video aesthetic quality assessment[J]. IEEE Transactions on Multimedia, 2020, 22(10): 2623-2634.

[5]GU J J, SU T, WANG Q H, et al. Multiple moving targets surveillance based on a cooperative network for multi-UAV[J]. IEEE Communications Magazine, 2018, 56(4): 82-89.

[6]FRATTOLILLO F, BRUNORI D, IOCCHI L. Scalable and cooperative deep reinforcement learning approaches for multi-UAV systems: A systematic review[J]. Drones, 2023, 7(4): 211.

[7]MINSKY M. Steps toward artificial intelligence[J]. Proceedings of the IRE, 1963, 49(1): 8-30.