Learning Robust Policy for Multi-UAV Collision Avoidance via Compact Causal Feature

Zhun Fan, Gaofei Han, Che Lin, Wenji Li, jie xu, Jiafan Zhuang

Keywords: Multi-UAV Systems, Collision Avoidance, Deep Reinforcement Learning, Representation Learning

TL;DR: We propose a Compact Causal Feature Learning framework to improve generalization and robustness of multi-UAV collision avoidance in unseen scenarios.

Abstract: Deep reinforcement learning (DRL)-based multi-UAV collision avoidance methods often exhibit limited generalization when deployed in unseen environments, primarily due to the reliance on non-causal and redundant visual features. Such overfitting to spurious correlations compromises both robustness and safety during real-world deployment. To address these limitations, this study proposes a novel Compact Causal Feature Learning (CCFL) framework that enables UAVs to learn compact and generalizable causal representations. Specifically, a Causal Feature Identification module is designed to disentangle input representations into causal and non-causal components, ensuring that the learned features preserve true environmental causality. Furthermore, a Redundancy Feature Compression module is introduced to remove redundant dependencies and compact the causal subspace, thereby enhancing generalization to previously unseen scenarios. Extensive experiments on a challenging UAV collision avoidance benchmark demonstrate that CCFL achieves substantial performance gains over state-of-the-art baselines, increasing individual success rates by 42.0\% and swarm success rates by 61.6\%. These results validate the effectiveness of compact causal feature learning for improving the adaptability, robustness, and safety of autonomous UAV systems operating in complex dynamic environments.

Area: Learning and Adaptation (LEARN)

Generative A I: I acknowledge that I have read and will follow this policy.

Submission Number: 385