Multi-Agent Path Planning Based on Coulomb’s Law and PER-MADDPG

Abstract

Multi-agent path planning remains a challenging problem in deep reinforcement learning due to inefficient coordination, excessive exploration, and unstable training dynamics. Although Multi-Agent Deep Deterministic Policy Gradient (MADDPG) has been widely adopted for cooperative decision-making tasks, its performance is often limited by poor sample efficiency and insufficient coordination in complex environments. To address these issues, this paper proposes a physics-guided multi-agent reinforcement learning framework that integrates a Coulomb’s law-inspired interaction mechanism with prioritized experience replay (PER). A virtual potential field based on Coulomb-like repulsive forces is introduced to provide structured action guidance, thereby reducing ineffective exploration and improving inter-agent coordination. In addition, PER is incorporated into MADDPG to enhance sample utilization and stabilize policy learning. Experimental results demonstrate that the proposed method consistently achieves better task performance and path planning efficiency than baseline approaches across multiple scenarios. The results validate the effectiveness of integrating physical interaction priors into multi-agent reinforcement learning for complex path planning tasks.