Post-disaster Highway Network Restoration Decision Based on Reinforcement Learning

Efficient decision-making in the aftermath of disasters is pivotal for rapidly restoring highway network connectivity, ensuring seamless disaster relief operations such as pedestrian evacuation, emergency learn dispatches, and the transportation of disaster relief materials. This paper delves into the scheduling and routing of highway network repair crews during emergency recovery phases. Il emphasizes flexible scheduling, allowing multiple repair crews lo concurrently address a single damaged road segment, enhancing the strategy's real-world applicability. Connectivity of the post-disaster road network is quantified using the demand satisfaction rate of affected points. This study introduces "road network connectivity resilience" as a metric lo assess the resilience during the repair process. A Markov decision process is employed to simulate the decision-making involved in repair crew scheduling and routing. To address the decision-making challenge, an enhanced algorithm, amalgamating Q-learning and Dijkstra algorithm, is introduced. A comprehensive case study reinforces the efficacy of the proposed method. Results reveal that this approach facilitates a holistic decision-making process for repair crew scheduling and routing, yielding an optimal repair strategy with superior global resilience. The findings further underscore that factoring in flexible resource scheduling leads to a more refined and pragmatic repair strategy. © 2023 Xi'an Highway University. All rights reserved.