Initial location assignment and dynamic reassignment optimization for urban evacuation guiders

Abstract

Urban evacuation is of great significance in saving human lives when disasters strike major cities. Comparing with the large population to evacuate, the number of evacuation guiders is much smaller. Since the location of evacuation guiders can greatly influence the evacuation process, it is crucial to identify the most critical locations to assign guiders. The assignment optimization for evacuation guiders usually suffers from partial information, partial control, uncertainty of individual behaviors and ever-changing traffic conditions during an evacuation. We divide this important problem into two parts: initial location assignment optimization and dynamic reassignment optimization. Both of them are considered in this paper and the major contributions are as follows. First, we mathematically model the assignment optimization problem using Markov decision process (MDP) framework. Second, we make a comparison among three methods that can be applied in initial location assignment optimization. Third, a simulation-based policy improvement method is developed to obtain optimal reassignment policies. Forth, some numerical examples are presented to demonstrate the performance of our methods. We hope this work sheds insight on evacuation guider assignment optimization problem in more general situations.