Approximate cutting plane approaches for exact solutions to robust optimization problems (Q2301927)

In this paper, approximate cutting plane approaches are proposed for robust optimization problems. Such an approach proceeeds iteratively in two stages: after the considered robust problem is solved with a reduced scenario set, a worst-case scenario is calculated which is afterwards added to the reduced scenario set. Then, again the robust problem is solved for the new reduced scenario set. Instead of solving both problems exactly, heuristics are proposed. It is proved that convergence to an optimal solution for such an approach is guaranteed under similar assumptions as for classical robust optimization cutting plane approaches. Experimental results are presented for robust mixed integer linear programs with mixed-integer polyhedral uncertainty sets.