An entropy model to measure heterogeneity of pedestrian crowds using self-propelled agents

Abstract

An entropy model to characterize the heterogeneity of a pedestrian crowd in a counter-flow corridor is presented. Pedestrians are modeled as self-propelled autonomous agents that are able to perform maneuvers to avoid collisions based on a set of simple rules of perception and action. An observer can determine a probability distribution function of the displayed behavior of pedestrians based only on external information. Three types of pedestrian are modeled, relaxed, standard and hurried pedestrians depending on their preferences of turn and non-turn when walking. Thus, using these types of pedestrians two crowds can be simulated: homogeneous and heterogeneous crowds.Heterogeneity is measured in this research based on the entropy in function of time. For that, the entropy of a homogeneous crowd comprising standard pedestrians is used as reference. A number of simulations to measure entropy of pedestrian crowds were conducted by varying different combinations of types of pedestrians, initial simulation conditions of macroscopic flow, as well as density of the crowd. Results from these simulations show that our entropy model is sensitive enough to capture the effect of both the initial simulation conditions about the spatial distribution of pedestrians in a corridor, and the composition of a crowd. Also, a relevant finding is that entropy in function of density presents a phase transition in the critical region.