Study on Uni-directinal Pedestrian Flow Based on Artificial Neural Network

Abstract

Casualties are caused in large-scale activities due to crowd and unavailable management, Thus investigation on pedestrian dynamics is of great significance to control pedestrian flow and reduce the casualty. Thus, based on tremendous data of occupants in a corridor, an uni-directional pedestrian flow model is formulated. This model is composed of Artificial Neural Network and pedestrian movement model. Based on big data, 25 factors are chosen as input values in Artificial Neural Network by considering 5-nearest-neighbor interaction pattern. The proposed model is tested for validation in respect of fundamental diagram, density distribution under the steady movement state and individual relationship between density and velocity. Simulated data is overlapped with experimental results and previous datasets: with density increasing, the velocity decreases nonlinearly; as for the microscopic study, the simulated results shows that greater density of individuals leads to smaller speed, which agrees with human characteristics in real life.