Modeling and simulation of pedestrian behaviors in crowded places

Abstract

Pedestrian simulation has many applications in computer games, military simulations, and animation systems. A realistic pedestrian simulation requires a realistic pedestrian behavioral model that takes into account the various behavioral aspects of a real pedestrian. In this article, we describe our work on such a model, which aims to generate human-like pedestrian behaviors. To this end, various important factors in a real-pedestrian's decision-making process are considered in our model. These factors include a pedestrian's sensory attention, memory, and navigational behaviors. In particular, a two-level navigation model is proposed to generate realistic navigational behavior. As a result, our pedestrian model is able to generate various realistic behaviors such as overtaking, waiting, side-stepping and lane-forming in a crowded area. The simulated pedestrians are also able to navigate through complex environment, given an abstract map of the environment.