A geometric cognition pedestrian dynamics model considering heterogeneous collision avoidance behaviors

Abstract

Collision avoidance is one of the crucial behaviors in pedestrian dynamics, especially in high-density scenarios where it frequently occurs to maintain crowd safety and stability. However, the heterogeneity of this behavior has not been well investigated. Given the advantages of geometric compatibility, a microscopic pedestrian flow model combining the Voronoi diagram and velocity obstacle approaches has been proposed to fully quantify heterogeneous collision avoidance behavior. In this model, the velocity region is dynamically partitioned into three sub-regions to describe the avoidance behaviors with different levels of aggressiveness. The concept of the difference to collision velocity is defined to characterize the evolution of aggressiveness, and a heuristic velocity region evolution rule is proposed to determine the walking velocity. To demonstrate the effectiveness of the model, a series of pedestrian flow scenarios are validated in terms of fundamental diagrams, self-organization phenomena, and behavioral heterogeneity. The impact of heterogeneous collision avoidance behaviors on crowd phenomena is then analyzed according to simulations. The DF-GLS algorithm is used to determine the crowd re-stabilization time. The results show that more aggressive behaviors lead to obvious stop-and-go wave and smaller Yamori’s band index of lane formation phenomenon. Sudden aggressive behaviors will disrupt crowd stability and even cause the crowd into a disorganized state. This model offers valuable insights and supports the optimization of emergency evacuation strategies with practical applications.