Developing a Pedestrian Agent Model for Analyzing an Overpass accident

Recent softwares have enabled us to apply pedestrian dynamics models into analyses on pedestrian accidents. The author had already developed a pedestrian dynamics model, called ASPF, based on the cell space model, that is evolved from Cellular-Automata (CA). After a couple of years’ efforts for continuous revision, we analyze the causes on an accident of Asagiri Pedestrian Overpass in 2001 even though retrospectively. In this paper, the typical existing models for pedestrian dynamics are reviewed, especially explaining on the class of cell space models. Next, ASPF (Agent Simulator of Pedestrian Flows) are explained, in that each pedestrian moves according to several behavioural rules on the cell-grid space of 40 cm side each. Based on not only the fundamental findings from the existing spatial researches but also these from the accident report, ASPF ver.2 is ‘tuned up’ carefully. ASPF ver.2 is to assess measures for managing pedestrian flows by focusing on the domino risk (density 3 to 5) that shows a symptom for the accident rather than a reconstruction of the real accident itself that had occurred at extremely high density (more than 10). The simulation results show that a two-way flow, combined with standing spectators (stoppers) can trigger an accident even on an overpass that satisfies present design standards. Moreover, we have confirmed that even simple traffic regulations such as partitions can be an effective measure to prevent a pedestrian accident.