Abstract: In order to study pedestrian behavior in crowded places in emergencies, a two-layer network model was proposed to simulate scenario of multiple paths in densely populated places. Dense layer and sparse layer were set in the model, in which traffic cost of sparse layer was lower. It is used to simulate path with shorter travel time in practice, and traffic cost in dense layer was higher, so it can simulate path with longer travel time in practice. The results show that critical capacity of system increases with increase of agents, and it is more sensitive to the number of agents in the sparse layer, and more pedestrians tend to choose sparse layer to pass. Occurrence of emergencies will lead to reduction of critical capacity of system. Meanwhile, interruption and degradation accidents in the sparse layer are more likely to cause the overall congestion. Congestion can be avoided by controlling flow into crowded places below critical value and avoiding accidents on shorter routes as far as possible.

Key words: emergencies, multi-path, moving behavior, crowded place, network model