Agent-based Simulation of Pedestrian Evacuation in Virtual Geographic Environment

Agent-based model has found great interest in simulation of pedestrian evacuation since it better represent the intelligent and self-organized behavior patterns of pedestrians. However, model validation and the user’s cognitive perspective have always been the two challenges in agent-based simulation. Therefore, this paper presents an approach to implement agent-based simulation of pedestrian evacuation in a virtual geographic environment (VGE) allowing interactions between avatars and agents, which extends the conventional agent-based modeling by adding human dimension. To test the approach, a pilot experiment has been carried out based on OpenSimulator. Since the avatars would perform close-to-real-life behaviors when a high sense of presence is achieved, such a study is useful for designing and validating agent-based simulation of pedestrian evacuation and helpful to understand pedestrian behavior from a first-person perspective.