Data-driven distributionally robust capacitated facility location problem

Ahmed Saif, Erick Delage

Abstract: Highlights•A distributionally robust version of the Capacitated Facility Location Problem is studied.•The distributional ambiguity set is constructed based on a Wasserstein ball centered at an empirical distribution and has a bounded support set.•Two exact iterative algorithms, based on column generation, are developed to solve the two-stage problem.•Conservative approximations and relaxations are proposed for the single- and two-stage problems.•Extensive numerical experiments are conducted to evaluate the computational performance of the algorithms and the quality of approximations.