Yunhe Tong - University of Bristol, Dept. of Engineering Mathematics
In emergencies, evacuations move pedestrians to safety. Extensive research on pedestrian exit choice assumes that pedestrians choose routes based on specific strategies to minimise the total evacuation time or to maximise the number of evacuees reaching safety within a given time frame. However, these objectives can lead to unfair allocations of evacuation resources for disadvantaged pedestrians who should be given priority, such as people with impaired mobility or ones who are nearer danger, such as fires. In this contribution, we propose a priority strategy where the defined vulnerable pedestrians can be preferentially assigned to exits in the case of limited exit resources. We formalise it in the social force model and illustrate its possible effects on pedestrian dynamics in three scenarios. First, normal scenarios where pedestrians have the same speed and being a long distance from exits is considered to be a vulnerability. Second, scenarios where pedestrians have different speeds where some pedestrians are vulnerable due to their lower speed. Third, fire scenarios where pedestrian close to the fire are regarded as vulnerable people. Our simulation results indicate that compared to the nearest exit strategy, the priority strategy can relieve crowd congestion to a certain extent but lead to a higher average velocity and longer walking distance. However, which strategy increases evacuation efficiency significantly depends on the initial distribution of pedestrians. We found machine learning methods including decision trees and support vector machines can predict well which strategy is better based on the initial distribution of pedestrian positions in normal scenarios, but the accuracy decreases in more complex scenarios. In addition, the increased crowd size and proportion of vulnerable pedestrians somewhat enhance the advantages of the priority strategy. Our work reveals the importance of appropriate allocation of priorities in evacuation and may thus be helpful for crowd safety.