Adaptive relief logistics for predictable disasters: Integrating professional vehicles with flexible crowdshipping

This study investigates a new relief logistics planning problem for predictable natural disasters with an actionable warning window. It integrates professional vehicles with crowdshipping and combines centrally pre-positioned inventories with pre-contracted suppliers to improve distribution flexibility and supply reliability. For the problem, we propose a novel multi-period distributionally robust chance-constrained model that accounts for dynamic and random demands, crowdshipping capacity, and traffic conditions to minimize the total relief cost under the worst-case scenario. Strategic decisions on facility location and supplier selection are made pre-warning, while post-warning tactical decisions involve multi-period relief supply distribution and vehicle assignment. To address computational challenges arising from the model's nonlinearity and large problem size, we reformulate it as a mixed-integer linear program and develop an improved column-and-constraint generation algorithm. The proposed algorithm features a customized three-phase framework consisting of decompose-and-relax, explore-and-repeat, and refine-and-repeat. It accelerates convergence via problem relaxation, scenario filtering, and iterative optimality tightening, with theoretical guarantees of finite-step convergence. Experimental results from a case study on historical typhoons in Shenzhen City, China, and randomly generated instances demonstrate the strong performance of the proposed model and algorithm. Insights relevant to management are also derived.