arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.04056cs.RO
收藏

基于因果特征选择的多无人机避碰鲁棒策略学习

Robust Policy Learning for Multi-UAV Collision Avoidance with Causal Feature Selection

College of Engineering, Shantou University(汕头大学工程学院);School of Computer Science, Guangdong University of Technology(广东工业大学计算机学院);University of Electronic Science and Technology of China(电子科技大学);International Cooperation Base of Evolutionary Intelligence and Robotics(进化智能与机器人国际合作基地)

Jiafan Zhuang, Gaofei Han, Zihao Xia, Boxi Wang, Wenji Li, Dongliang Wang, Zhifeng Hao, Ruichu Cai, Zhun Fan

AI总结:

针对深度强化学习的无人机避碰导航方法泛化性差的问题,提出可集成到策略网络的因果特征选择模块,过滤非因果因素以降低虚假关联影响,在未知场景下实现了更优的鲁棒避碰性能。

摘要:

在未知复杂户外环境中,无人机集群的避碰导航是一项极具挑战性的问题,要求无人机能够穿越各类障碍物与复杂背景。现有基于深度强化学习(DRL)的避碰导航方法虽展现出不错的性能,但泛化能力较差,会在未知环境中出现性能下降。为解决该问题,我们探究了DRL泛化能力弱的成因,提出了一种新型因果特征选择模块,该模块可集成到策略网络中,有效过滤表征中的非因果因素,从而降低非因果因素与动作预测间虚假关联的影响。实验结果表明,所提方法可实现鲁棒的导航性能与有效避碰,尤其在背景和障碍物均未知的场景下,性能显著优于现有最优算法。

英文摘要:

In unseen and complex outdoor environments, collision avoidance navigation for unmanned aerial vehicle (UAV) swarms presents a challenging problem. It requires UAVs to navigate through various obstacles and complex backgrounds. Existing collision avoidance navigation methods based on deep reinforcement learning show promising performance but suffer from poor generalization abilities, resulting in performance degradation in unseen environments. To address this issue, we investigate the cause of weak generalization ability in DRL and propose a novel causal feature selection module. This module can be integrated into the policy network and effectively filters out non-causal factors in representations, thereby reducing the influence of spurious correlations between non-causal factors and action predictions. Experimental results demonstrate that our proposed method can achieve robust navigation performance and effective collision avoidance especially in scenarios with unseen backgrounds and obstacles, which significantly outperforms existing state-of-the-art algorithms.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑