Partitioning Vertical Evacuation Areas in Umeda Underground Mall to Minimize the Evacuation Completion Time

Abstract

When an underground mall is flooded, the shoppers should be evacuated to a building connected to the mall. However, the number of evacuees from a large-scale underground mall will exceed the capacity of the evacuation center. Furthermore, the evacuation time may be delayed. This paper proposes a mathematical programming problem that minimizes the evacuation completion time on a general planar graph of a partitioned evacuation area with a specified sink capacity. We also propose a workflow for translating the general geometric spatial data to graphical data. The problem is applied to the real spatial data and evacuation setting of Umeda underground mall in Osaka, Japan. The problem's performance is compared with that of the conventional problem that minimizes the total evacuation distance, and its accuracy is confirmed in a multi-agent simulation. The validity of the proposed method is also discussed.