Herding in Real Escape Panic

Abstract

We introduce the element of copying in an agent-based model of escape panic to describe with greater accuracy the exit behavior of mice that are escaping from a flooded two-exit chamber. Aside from the panic threshold ϕ (0 ≤ ϕ ≤ 5), our model utilizes the imitation tendency α (0 ≤ α ≤ 1) such that agents with ϕ = 0, are calm and tend to stay put while those that are likely to copy their neighbors are described by large α values. A high degree of copying among escaping agents favors the emergence of herding behavior. Both the Moore and the von Neumann neighborhood are tried to depict the movement of agents in a plane. Herding decreases the exit throughput Q by causing an inefficient utilization of the two available exits for escape. The dependence of Q with α and the exit door separation are highly nonlinear. The inclusion of α has significantly improved the capability of our model to explain the Q-behavior that was observed in the mice experiments. Interestingly, simulation results show that copying could promote faster room evacuation at α ≈ 0.5 and especially at high room occupancy rates (> 60%). At α ≈ 0.5, an agent is equally likely to copy or ignore the action of its neighbor.