Using a Multi-Scale Model for Simulating Pedestrian Behavior

Abstract

In order to model realistic pedestrian crowds, different aspects on different scales have to be taken into account. Besides behavioral aspects, locomotion on short-scale and human navigation on large-scale have to modeled appropriately. In the simulation models existing to date, these two aspects are modeled separately. To overcome the limitations of currently available models, this paper presents a new hybrid multi-scale model, which closely links information between the short-scale and the large-scale layer to improve the navigational behavior. In the presented hybrid navigation model, graph-based methods using visibility graphs are used to model large-scale wayfinding decisions. The pedestrians’ movements between two nodes of the navigation graph (the short-scale) are modeled by means of a dynamic navigation floor field. The floor field is updated dynamically during the runtime of the simulation, explicitly considering other pedestrians for determining the fastest path.