Exact methods for discrete {varGamma }-robust interdiction problems with an application to the bilevel knapsack problem

Abstract

Developing solution methods for discrete bilevel problems is known to be a challenging task—even if all parameters of the problem are exactly known. Many real-world applications of bilevel optimization, however, involve data uncertainty. We study discrete min-max problems with a follower who faces uncertainties regarding the parameters of the lower-level problem. Adopting a varGamma -robust approach, we present an extended formulation and a multi-follower formulation to model this type of problem. For both settings, we provide a generic branch-and-cut framework. Specifically, we investigate interdiction problems with a monotone varGamma -robust follower and we derive problem-tailored cuts, which extend existing techniques that have been proposed for the deterministic case. For the varGamma -robust knapsack interdiction problem, we computationally evaluate and compare the performance of the proposed algorithms for both modeling approaches.