Data mining for pedestrian movement in large-scale experiments

Abstract:
In order to study the essence of pedestrian flow under high densities, we organized one large-scale pedestrian flow experiment, and the extreme density as high as 9ped/m2 was reached. To get more understandings, the microscopic pedestrian data in the experiment, including velocities and trajectories, are particularly studied by data mining technology in this paper. For uni-directional flow, we introduced the typical examples of hyper-congested regime and over-congested regime. For bi-directional flow, the “temporary deadlock” before the lane formation and the loop oscillation after the lane formation are discussed. The dynamics revealed from the microscopic results can be a great help for related computer simulations.