Accidental emergence within an agent based model: Simulation of agent interactions in an emergency situation

This paper describes an agent based modeling (ABM) study oriented towards simulating crowds and movements of agents within a confined area. The region of interest is a rectangular room with one or two exits, where the sounding of an alarm triggers agents to move towards the exit(s). The agent based model is built around a game framework with emphasis on visualization of agent behavior. Agent interactions are local, well defined and approximate behaviors modeled in other crowd simulations. Agents have six behavioral modes, which control the agents´ target(s), their capacity to make decisions regarding which path to take, and their ability to communicate to one another. The initial intent was to observe the behaviors during egress where crowding occurs at the exit bottleneck. Behaviors of agents leaving are typically as expected as all interactions were coded to precipitate the events. The accidental or serendipitous behavior was a consequence of extending the simulation beyond the observations of crowd formation associated with the exodus. It is the visualization afforded by a serious game development framework and visualization which facilitated the observation of emergent behavior. This accidental emergence or spontaneous order was observed and not explicitly controlled or programmed.