Abstract
Reducing damage on human society from natural disasters is an urgent matter of many cities and countries all over the world, and especially, preventing people from over-concentrating in urban areas after the disaster is important. Predictions of people movement after disasters can be utilized for making quick rescue plans which can avoid the concentrated areas, efficiently distributing supplies such as water, food and shelter, and preventing further damages caused by over-crowding. In this paper, we present our method of accurately predicting people's movement under disaster situations in a metropolitan scale. Our method combines multi agent simulation using a disaster behavioral model, and real-time observation data using Particle Filter method. We verified the effectiveness of our method by experimenting it on the Great East Japan Earthquake, and obtained a high prediction accuracy. Also, the number of people bound for each destination matched the survey results carried out by a different research, supporting the accuracy of the method.
