Agent-based passenger modeling for intelligent public transportation

This study focuses on developing and illustrating a passenger motion and behavior model for public transportation applications within Intelligent Transportation System research. The specifics of the model are selected to fit the needs of public transportation examples, as opposed to the more generic pedestrian models investigated in the literature. The approach is agent based-agents being modeled using hybrid state machines, and each high level state corresponding to a mode of behavior. At different behavioral modes, different social forces-including attractive/repulsive forces from destinations, people, and obstacles-are active in determining pedestrian motions. In our pedestrian model, information regarding the situatedness of agents-e.g. door locations, crowdedness of areas, other pedestrians within peripheral vision, approaching flow from opposing direction within the central vision, obstacles such as walls-are incorporated at both high-level behaviors and low-level pedestrian motion control. The structure of our model also allows for future expansions. The developed model is shown to be useful in a number of examples-where the emergent collective behavior of pedestrians and the resulting passenger loading/unloading dynamics are studied. The simulation results demonstrate the utility of our approach.