Multi-UAV Cooperative Obstacle Avoidance Path Planning Based on PER-MASAC

To address the challenges of limited local perception and high decision-making coupling in multi-UAV cooperative obstacle avoidance under dynamic and complex battlefield environments, this paper proposes a Multi-Agent Soft Actor-Critic algorithm integrated with Prioritized Experience Replay (PER-MASAC). The method constructs a multidimensional state space incorporating target navigation, obstacle perception, and neighbor coordination. A hybrid reward function combining sparse and shaped rewards is designed to enhance learning efficiency. Furthermore, a SumTree-based priority sampling mechanism is introduced to improve the reuse of high-value experiences. Experimental results in a high-fidelity simulation environment demonstrate that compared to MASAC and other mainstream algorithms, PER-MASAC achieves significantly better performance in terms of convergence speed and task success rate. This provides a verifiable solution for intelligent decision-making of UAV swarms in unknown and complex environments.