Integrated planning of casualty transportation and medical supply distribution in the post‐disaster relief network: a Q‐learning‐based multi‐objective robust optimization approach

In recent years, natural disasters have become more frequent and destructive, causing major infrastructure damage and significant loss of life. The unpredictable number of casualties and their rapidly changing conditions after a disaster pose significant challenges to the post‐disaster relief efforts. Considering the uncertain number and dynamic condition of the casualties, we investigate an integrated relief planning problem of emergency facility locations, casualty transportation routes, and medical supplies distribution. A robust optimization approach is employed, and the concept of dynamic casualty death probability resulting from the treatment delays is introduced. Then, a tri‐objective optimization model is formulated to minimize the casualty death probability, the latest arrival time of medical supplies, and the total operational costs. To solve the model, a multi‐objective particle swarm optimization algorithm is proposed, which is enhanced with Q‐learning to guide the search direction by dynamically adjusting the inertia weights. Moreover, a local search mechanism is incorporated to improve the population diversity, resulting in a Q‐learning‐based multi‐objective particle swarm optimization algorithm with local search. The effectiveness of the proposed algorithm is validated through a set of numerical experiments. A case study of the Yushu Earthquake further demonstrates the model's practical applicability.