Improving pedestrian micro-simulations with event steps

Abstract

Microscopic pedestrian models describe individual pedestrian behavior and the interaction of pedestrians with other pedestrians and obstacles. Continuous time models generally calculate the acceleration of pedestrians due to repulsive or attractive interactions using for instance distances to other pedestrians and obstacles within a two-dimensional influence area. Two problems that usually arise with these types of models when simulating very large crowds are extreme accelerations that occur due to very short distances to other pedestrians and obstacles and large computational times to assess and to calculate the accelerations for individual pedestrians. This paper presents a hybrid pedestrian management algorithm that combines a traditional optimized time-based simulation and an event-driven simulation. This way, the task of assessing the surroundings and the task of dealing with interactions on very short distances are each treated in an optimized way leading to more reliable accelerations in high densities as well as shorter calculation times. © 2009 Springer-Verlag Berlin Heidelberg.