Abstract: Existing researches on pedestrian flows have mainly focused on evacuation optimization during or after emergencies, with little attention paid to crowd management before emergencies. This research formulates a simulation model of pedestrian flow based on a multi-agent system, with the utility that each pedestrian gets in the model defined on the distance to the exit and the density of the cell. Then, Bayesian-Nash Equilibrium is employed to search for the target cell of maximum expected utility. After that, the model is validated by a real scenario and is found to have a good consistency. At last, with the experimental data collected from the different scenarios and the walking preferences taken into consideration, this paper reaches the conclusion that the safe flow rate increases by about 3ped/s as the tunnel width expands by 1m. The model proposed in this paper is proved to be working very well on dense pedestrian flows, and the conclusion would provide reference for the management of dense pedestrian flows.

Key words: tunnel pedestrians, Bayesian-Nash Equilibrium, simulation of pedestrian flow, safe flow rate