A heterogenous parallel algorithm for stadium evacuation route assignment

Abstract

To speed up the time-intensive, long-duration calculation process of large-scale, multi-criteria evacuation route assignment problem, this paper introduce a GPU-accelerated genetic algorithm which utilize the lightweight many-thread architecture of CUDA GPU to fasten the evaluation procedure of evacuee movement. The proposed algorithm introduces a two-level heterogeneous parallel model to harness the power of CPU and GPU separately. Experimental results show that the proposed algorithm run 6.23 times faster than sequential version counterpart and can achieve better search result within specified time unit with respect to Wuhan sports center evacuation route assignment problem. © 2013 Springer-Verlag.