Evacuation trees with contraflow and divergence considerations

In this chapter, we investigate the well-studied evacuation planning problem, in which vehicles use the available road transportation network to reach safe areas (shelters) in the face of an upcoming disaster. To successfully evacuate these vehicles located in danger zones, our evacuation process needs to be fast, safe, and seamless. We enable the first two criteria by developing a macroscopic, time-dynamic evacuation model that aims to maximize a reward function, based on the number of people in relatively safer areas of the network at each time point. The third criterion is achieved by constructing an evacuation tree, where vehicles are evacuated using a single path to safety at each intersection. We extend on the definition of the evacuation tree by allowing for divergence and contraflow policies. Divergence enables specific nodes to diverge their flows into two or more streets, while contraflow allows certain streets to reverse their direction, effectively increasing their capacity. We investigate the performance of these policies in the evacuation networks obtained, and present results on two benchmark networks of Sioux Falls and Chicago.