Algorithmic Evaluation of Fire Evacuation Efficiency Under Dynamic Crowd and Smoke Conditions

This study developed a fire evacuation simulation model for a six-level underground station to evaluate evacuation efficiency under both dynamic and static conditions, including structural damage, smoke propagation, and real-time crowd congestion. Two representative pathfinding algorithms, Dijkstra’s and A*, were applied to analyze evacuation performance across eight fire scenarios occurring at different locations within the station. When only static factors were considered, both algorithms yielded identical maximum evacuation times, indicating comparable performance. However, the A* algorithm exhibited a significantly shorter computation time than Dijkstra’s, demonstrating higher operational efficiency. When dynamic variables such as real-time congestion and smoke-induced visibility reduction were introduced, the maximum evacuation times varied irregularly between the two algorithms. This outcome suggests that, under dynamic fire conditions, route guidance based solely on current information rather than predictive modeling may lead to suboptimal evacuation outcomes. Therefore, this study emphasizes the importance of establishing a predictive disaster management system capable of forecasting fire and smoke propagation, as well as a centralized control system that can dynamically distribute evacuees to enhance evacuation efficiency in deep underground stations.