Characteristic time based social force model improvement and exit assignment strategy for pedestrian evacuation

Abstract

Pedestrian modeling is essential for evacuation simulations. One unrealistic phenomenon observed in AnyLogic and open-source software is that running pedestrians often hit the wall then they try to pass an exit. We find that there is only one characteristic time in the social force model (SFM), which means SFM does not divide the desired speed effect on the normal and tangential direction of current speed. To avoid hitting the wall, we propose to calibrate the tangential characteristic time that changing moving direction with real-life experiment. Comparisons of simulation and real experiment show that this characteristic time is 0.2 s. Based on this improvement, another common problem of existing evacuation strategies is studied. The problem is that they only assume exit capacity (EC) of gate is linearly equal to door width. As such, an EC based exit assignment strategy is proposed to fully use multiple gates’ capacities. Furthermore, we propose to regard congested bottlenecks as virtual gates (VG), which is equivalent to real gates in EC computation. Simulation results show that our nonlinear EC based assignment strategy outperforms other methods because it can compute the queue time near exit more accurately.