VR-based fire evacuation in underground rail station considering staff’s behaviors: model, system development and experiment

As an enclosed and normally crowded environment, underground rail station is a dangerous place when fire emergency happens. Traditional methods usually adopt simulation or drill to evaluate station design or contingency plans, which differ far from real situations in both parameters and people’s complicated behavior in an emergency, especially the staff’s guiding action. Staff’s behaviors are enormously important for evacuation, because evacuation path is complex and smoking diffusing is in the same direction with the path to some extent. In this regard, a fire evacuation system considering staff’s behavior in virtual reality environment was developed. The distinguished features of the system embodied in three aspects. First, it provided a vivid, interactive and immersed platform for staff to practice and optimize evacuation plan. Second, staff’s behaviors and their implication on passengers’ evacuation were modeled. Passengers were treated as multi-agent that can feel the surrounding environment including staff’s guiding, and they can change their paths in real-time which implemented by a dynamic path choice model. Third, input and visualization of the fire data extracted from fire simulation, and the 3D station model, provide quite real evacuation environment. Recruiting staff from operation company as subjects, the platform was applied for fire evacuation test considering staff’s intervention. The results demonstrate that the proposed model and developed system can be effectively used for contingency plan evaluation and improvement.