Determining Optimal Evacuation Decision Policies For Disasters

Abstract

Decision making in the face of uncertainty is a difficult task, and this is exacerbated when the decision is irreversible, it involves a near-term deadline, and/or the cost of a bad decision is high. Deciding whether to stay or evacuate from an impending natural disaster is difficult for all of these reasons. This thesis explores the evacuation decision as a Markov decision problem. We develop a generic disaster model to explore the tensions and tradeoffs in the decision to evacuate and use a dynamic programming algorithm to determine optimal decision policies for the decision maker. We explore how these policies are affected by evacuation costs as well as disaster uncertainty.