Agent-based Dynamic Model For Pedestrian Counter Flow

Abstract

This paper presents the establishment of the Agent-based Dynamic Model for Pedestrian counter flow (ADMP). It was shown that a rule set of pedestrian behaviors and its combination with a dynamic model for pedestrian flow was capable of effectively capturing the crowd flow rules of pedestrians using agent-based technology. Two digital tests of bottlenecks of evacuation passage in counter flow are presented, in which 200 and 300 pedestrians were stochastically distributed in a space measuring 40000 x 10000 (mm). Crowd behavior phenomena, such as arching, jamming, lane forming, and oscillating could be observed after certain critical points. From a comparison of the two digital tests, it could be seen that, with an increase in pedestrian density, the counter flow became more and more disorderly, and the two different directions of flow struggled more excitedly as a result. It illustrated that the ADMP was efficient in the simulation of pedestrian flow and is a useful tool for fire safety research and management.