Prioritizing Postdisaster Recovery of Transportation Infrastructure Systems Using Multiagent Reinforcement Learning

Postdisaster reconstruction of transportation infrastructures generally entails complex and multiobjective planning and implementation options under uncertainty because of a large number of underlying subjective and objective factors, including social, economic, political, and technical aspects. With limited federal, state, and local resources, it is also challenging for decision-makers to establish a meticulous plan for postdisaster transportation recovery. However, previous studies mainly dealt with the specific planning or execution part of the postdisaster recovery process and rarely considered a comprehensive set of objectives in their investigations. This paper aims to develop a new prioritization approach for rapid and optimized postdisaster recovery that evaluates recovery priorities of damaged transportation infrastructure systems and affected regions through a multiagent system using a reinforcement learning technique. The proposed model contributes to the body of knowledge by providing a new optimization framework, considering transportation network recovery, and minimizing the social impact of the current prolonged recovery process on affected communities. This new methodology is expected to help public agencies make an informed decision for distributing given resources and structurally arranging disaster recovery processes of transportation systems by simulating real-world high-dimensional disaster scenarios and optimizing their recovery plans. In particular, the proposed approach pursues to assist disaster-relevant practitioners in considering a holistic perspective for comprehensive decision-making, incorporating diverse factors of planning transportation recovery and assigning their resources according to socioeconomic factors of affected communities.