Reinforcement Learning Based Recovery of Education After Natural Hazards

School facilities are essential in offering a safe and stable environment for education. However, schools are prone to damage during natural hazard events, causing education disruptions. It is imperative to rapidly reopen schools after such events so that students can receive timely education. In the meantime, the continued delivery of education services also depends on the infrastructure supporting schools, such as roads and bridges. In this paper, a reinforcement learning-based approach to recovering schools considering associated bridges is proposed. Two separate but interactive crews are modeled to deliver recovery actions to schools and bridges respectively. The effectiveness of the proposed approach is demonstrated with a case study in the Dominican Republic. Results show that the use of reinforcement learning leads to the identification of recovery strategies that are more efficient than the ones identified by heuristic based ranking priorities.