Mean-Field Games Modeling of Anticipation in Dense Crowds

TL;DR

This paper develops a Mean-Field Game model to simulate dense crowd dynamics, incorporating anticipation and information processing limitations, and validates it against experimental data showing effective capture of crowd behaviors.

Contribution

It introduces a novel MFG model with a discount factor for anticipation, applied to dense crowds crossing an intruder, and demonstrates its effectiveness through experimental comparison.

Findings

Model accurately reproduces anticipatory crowd behaviors

Incorporating anticipation improves simulation realism

Effective in different pedestrian orientations

Abstract

Understanding and modeling pedestrian dynamics in dense crowds is a complex yet essential aspect of crowd management and urban planning. In this work, we investigate the dynamics of a dense crowd crossed by a cylindrical intruder using a Mean-Field Game (MFG) model. By incorporating a discount factor to account for pedestrians' limited anticipation and information processing, we examine the model's ability to simulate two distinct experimental configurations: pedestrians facing the obstacle and pedestrians giving their back to the intruder. Through a comprehensive comparison with experimental data, we demonstrate that the MFG model effectively captures essential crowd behaviors, including anticipatory motion and collision avoidance.