A Multi-Agent and PSO Based Simulation for Human Behavior in Emergency Evacuation

A new emergency evacuation model based on Multi- Agent framework and Particle Swarm Optimization was presented. This model simulated human´s individual behaviors and social behaviors in multi-exit evacuation environment. The Linear Weight Decreasing Particle Swarm Optimization (LWDPSO) was introduced to simulate individual´s movement. A hierarchy of behavior rules was described, and a series of individual behaviors and social behaviors during evacuation were defined. A prototype system of emergency evacuation simulation was implemented based on Geographic Information System (GIS) application framework. The result shows that the simulation system well performs some typical evacuation behaviors, the model based on multi- agent and LWDPSO has good efficiency and practicability.