A Computational Framework for Egress Analysis with Realistic Human Behaviors

Studies of past emergency events indicate that evacuating occupants often exhibit social behaviors that affect the evacuation process. This paper describes a multi-agent based simulation tool which enables the modeling of social behaviors during evacuation. In this modeling framework, each agent has a three-level representation that allows users to incorporate individual, group, and crowd behavioral rules in simulation. This paper describes the basic framework and the implementation of several social behaviors, which are based on recent social science studies about human response in emergency situations. Simulation results from the prototype reveals that social behaviors exhibited by the evacuating crowd can lead to changes in the overall egress time and pattern. By representing the virtual agents and the environment specific to evacuation situation, the research addresses the issues in incorporating human and social behaviors in egress simulation.