[4F2-GS-10m] AI応用：移動と交通

[4F2-GS-10m-04] Evaluation of Security and Efficiency in Large Crowd Events

〇Shusuke Shigenaka^1,2, Shunki Takami^1,2, Masaki Onishi² (1. University of Tsukuba, 2. Artificial Intelligence Research Center, AIST)

Keywords:multi-criteria evaluation, pedestrian simulation, evolutionary algorithm

In this study, we proposed a crowd control index to evaluate the efficiency of crowd control strategies for safety in a large-scale event. These control strategies are essential to manage mass gatherings appropriately and to prevent crowd-induced accidents. We first defined the crowd control index in terms of the efficiency and the safety level of the control, based on traffic flow parameters. Then, the index was compared with the existing index by using four types of headcount distribution, and 19683 control patterns. To simulate these patterns in an evacuation experiment, we used a pedestrian simulator. In the experiment, we set two types of pedestrian distribution, 30000 and 45000 pedestrians each, and two patterns of departure distribution using a large-scale outdoor festival in Japan. The experimental results show that the proposed index is more stable than the existing index to evaluate the efficiency of the crowd control patterns.