Title:Solving Decision-Dependent Robust Problems as Bilevel Optimization Problems

Abstract:Both bilevel and robust optimization are established fields of mathematical optimization and operations research. However, only until recently, the similarities in their mathematical structure has neither been studied theoretically nor exploited computationally. Based on the recent results by \textcite{goerigk2025}, this paper is the first one that reformulates a given strictly robust optimization problem with a decision-dependent uncertainty set as an equivalent bilevel optimization problem and then uses solution techniques from the latter field to solve the robust problem at hand. If the uncertainty set can be dualized, the respective bilevel techniques to obtain a single-level reformulation are very similar compared to the classic dualization techniques used in robust optimization but lead to larger single-level problems to be solved. Our numerical study shows that this leads to larger computation times but may also slightly improve the dual bound. For the more challenging case of a decision-dependent uncertainty set that cannot be dualized because it is represented as a mixed-integer linear problem, we are not aware of any applicable robust optimization techniques. Fortunately, by exploiting the corresponding bilevel reformulation and recent bilevel solvers, we are able to present the first numerical results for this class of robust problems.