Statistical Models for Pedestrian Behaviour in Front of Bottlenecks

Abstract

Understanding the movement of human crowds is important for our general understanding of collective behaviour and for applications in building design and event planning. Here, we focus on the flow of a crowd through a narrow bottleneck. We develop statistical models that describe how pedestrian behaviour immediately in front of a bottleneck affects the time lapse between consecutive pedestrians passing through the bottleneck. With this approach, we isolate the most important aspects of pedestrian behaviour from a number of candidate models. We fit our models to experimental data and find that pedestrian interactions immediately in front of the bottleneck appear to be less important for the observed time lapses than interactions further away from the bottleneck. Furthermore, we demonstrate how our approach can be used to rigorously compare microscopic pedestrian behaviours across different contexts by fitting the same statistical models to three separate datasets. We suggest that our approach is a promising tool to establish similarities and differences between simulated and real pedestrian behaviour.