Exploiting Spatial-temporal Heterogeneity for Agent-based Simulation of Pedestrian Crowd Behavior

Agent-based simulations of pedestrian crowd behavior include a large number of agents, each of which typically decides its movement behavior in a time-based manner. Such a time-based approach makes it difficult to exploit the crowd system’s spatial and temporal heterogeneity resulting from agents’ non-uniform movements. This paper presents a discrete event approach for simulating pedestrian crowd. In the discrete event approach, an agent’s decision making is triggered by the “changes” of its external environment and/or its own internal states. This discrete event model naturally track the crowd system’s “activities” in both space and time, and represents a different way of agent modeling in virtual worlds. In this paper, we describe the discrete event agent model, present experiments results, and compare the performance between the discrete event approach and discrete time approach for different space resolutions.