Abstract
Given maps of an evacuee population, shelter destinations and a transportationnetwork, the goal of intelligent shelter allotment (ISA) is to assign routes, exits and shelters to evacuees for quick and safe evacuation. ISA is societally important due to emergency planning and response applications in context of hazards such as floods, terrorism, fire, etc. ISA is challenging due to conflicts between movements of evacuee-groups heading to different shelters and transportation-network choke-points. State of the practice based on Nearest Exit or Shelter (NES) paradigm addresses the former challenge but not the latter one leading to load-imbalance and slow evacuation. Recent computational development, e.g., capacity-constrained route planning (CCRP), address the latter challenges to speedup evacuation, but do not separate evacuee groupsgoing to different shelter destinations. To address these limitations, we propose a novel approach, namely, Crowd-separated Allocation of Routes, Exits and Shelters (CARES) based on the core idea of spatial anomaly avoidance. Experiments and Hajj case study (Makkah) show that CARES meets both challenges by providing much faster evacuation than NES and much lower evacuee-group movement-conflicts than CCRP.