Multi-Agent Based Simulation of Passenger Evacuation Behaviors and Emergency Organization Optimization in Subway Fire Scenarios

Subway fires present serious risks to passenger safety because of dense crowds, confined spaces, and limited evacuation routes. This study develops a multi-agent-based simulation framework to analyze passenger evacuation behaviors and optimize emergency organization strategies in subway fire scenarios. Passengers, emergency personnel, and environmental hazards are modeled as autonomous agents with heterogeneous characteristics, including panic levels, movement abilities, and route selection preferences. The simulation dynamically incorporates smoke diffusion, visibility reduction, and temperature changes to reproduce realistic evacuation conditions. Results show that adaptive emergency guidance and dynamic exit management can effectively reduce evacuation time, mitigate congestion, and improve overall evacuation safety, providing valuable support for intelligent subway emergency management and resilient evacuation planning.