A Newly developed Mesoscopic Model on Simulating Pedestrian Flow

Abstract

This paper presents a newly developed mesoscopic model with grid-based structure for building evacuation simulation. The evacuation space in this model is discretized into cells with larger size than that of the microscopic models. This mesoscopic model directly computes the density flow between the cells governed by the density flow algorithm. The computation speed is higher than the traditional cellular automata model in which the movements of all pedestrians should be tracked for calculating the crowd density. To test the feasibility of this model, we applied it to a typical scenario in which pedestrians evacuate from a square room with single exit and conducted parameter sensitivity study on the length of the time step δ. The simulation results show that, within a valid range, changing δ only has minor influence on the evacuation process. We can directly identify area with high density as bottleneck from dynamically changed density map even the relatively large time step is adopted. In addition, the commercial package AnyLogic was used to benchmark the performance of this model. The results show that the mesoscopic model discerns evacuation patterns in more details compared to the macroscopic models and with higher efficiently computational speed than the microscopic models.