Walking model on passenger in merging passage of subway station considering overtaking behavior

Abstract As bottlenecks of walking facilities in subway stations, merging passages are prone to congestion when passengers walk and evacuate under daily operation and emergency conditions. The general social force model (SFM) can accurately reproduce the self-organization phenomena of passenger crowds, such as the strip and lane formation. However, the SFM cannot model the overtaking behaviors of individuals who try to avoid collisions with the preceding walkers. To depict the walking behaviors of passengers in merging passages, the SFM is improved by considering the overtaking behaviors of passengers (OSFM) in this paper. Firstly, the trajectories of heterogeneous passengers are abstracted from video footages including the processes of overtaking behaviors. Next, the principle of overtaking, which leads passenger to detour with the shortest route and avoid colliding with the preceding passenger, is put forward to reproduce the overtaking processes of heterogeneous passengers in merging passages. The OSFM is verified to be more accurate than the social force model with only a centripetal force appended (ISFM). Then, the probability of overtaking behavior is estimated based on analysis of the simulation scenarios with various population densities. Finally, some simulation experiments of heterogeneous passengers with different arrival intervals in the merging passages are conducted based on the OSFM. The results show that when population density is below 0.08 ped/m2 or exceeds 1.56 ped/m2, the overtaking behavior scarcely occurs, but when the density is in the range of 0.25–0.4 ped/m2, the overtaking behaviors are mostly likely to occur. The probability of overtaking behavior will decrease with the density increasing when the density exceeds 0.4 ped/m2. What’s more, the total passing time and walking speeds are more sensitive to the change of layouts of merging passages when the arrival interval of passengers obeys exponential distribution.