Crowd evacuation path planning and simulation method based on deep reinforcement learning and repulsive force field

Hongyue Wang, Hong Liu, Wenhao Li

Abstract: Path planning is essential for simulating crowd evacuation. However, existing path planning methods encounter challenges, including unbalanced exit utilization, ineffective obstacle avoidance, and low evacuation efficiency. To address these issues, this paper presents a path planning method based on Deep Reinforcement Learning (DRL) and a Repulsive Force Field (RFF) for crowd evacuation simulation. First, a dynamic exit scoring mechanism is proposed and integrated into the DRL training process to balance exit utilization during evacuation. Additionally, we address the sparse reward issue in DRL by extracting key points from actual evacuation trajectories as short-term goals. Finally, we enhance the movement strategy output by constructing an RFF to improve obstacle avoidance in complex environments. Experimental results demonstrate that the proposed method effectively avoids obstacles and efficiently completes evacuation tasks.