Simulating Crowd Egress Dynamics Using Multiple Agents and Social Comparison Theory

Abstract

The understanding of crowd movement is important to planning and improving shared public places, not only to effectively and efficiently facilitate the comfortable movement of individuals, but also to guarantee the safety of individuals, especially under conditions of danger when quick and orderly evacuation of a mass of individuals is desired. Because data from real evacuation are hard to obtain and conducting replicate experiments on humans is ethically questionable, the characterization of crowd egress dynamics has been confined to simulation and modeling. We introduce a simulation approach that hybridizes multi-agent systems (MAS) with the social comparison theory (SCT) that provides the capability to simulate more human crowd phenomena than the more common social force model (SFM). SFM has been proven to show real world crowd phenonema such as the “faster-is-slower” in escape panic, “arching” and “bursty exit” as side effects to “clogging” on exit ways, “flocking,” “bidirectional lane formation,” and “roundabout formation.” Simulations using our MAS-SCT hybrid are able to exhibit all these phenomena and two more individual behaviors: (1) Imitation– where individuals tend to move in groups whose members they think would have the same opinion as theirs; and (2) Contagion– where people tend to “adopt” the behavior of others in the same group. Because of these, we propose that our MAS-SCT approach is more akin to modeling humans and real-world objects in very realistic ways, and thus can be used with higher confidence in performing what-if scenarios to aid decision makers, designers and researchers.