Stochastic Two-echelon Location-Routing

In the two-echelon location-routing problem one has to decide on the number and the location of primary warehouses as well as intermediate distribution platforms; while xing the capacity of the links between them. The system must be dimensioned to enable a ecient distribution of goods to customers under a stochastic and multi-period demand over a planning horizon. I.e., in the second echelon, the planning prescribes vehicle routes that visit ship-to-points from selected distribution platforms, minimizing the total expected transportation cost. For this two-stage stochastic program with recourse, we developed a Primal-Dual approach. We combine a Benders decomposition where the rst stage are the location and capacity decisions, with a Branch-and-Price algorithm to tackle the second stage routing problem. Our preliminary computational study illustrates the tight optimality gaps that can be obtained on realistic instances