Parallel computation using GPGPU to simulate crowd evacuation behaviors: Planning effective evacuation guidance at emergencies

Abstract

We propose parallel computing to simulate crowd evacuation behavior. It allows evacuation of ten thousands of agents to be simulated faster than does the existing system. Our prototype system consists of a new traffic simulator and scenario generator. The traffic simulation system uses a general purpose graphics processing unit (GPGPU) and simulates the agents' movements in a three-dimensional map. Our proposal enables realistic evacuation simulations and provides a platform that widens the applications of RoboCup Rescue Simulation to, for example, crowd evacuation from buildings. The evacuation simulations help security offices to prepare manuals for emergencies. © 2014 Springer-Verlag Berlin Heidelberg.