A Brief Overview of Interdiction and Robust Optimization

Abstract

Two-player optimization problems span an impressive array of possible situations, including cases in which both players optimize their own objective with no regard for the other’s goals, or in which one agent seeks to impede the other’s objective. The agents may commit their decisions simultaneously, using either deterministic or random (mixed) strategies. Alternatively, they can play them in sequence, where one agent has complete or partial knowledge of the other’s decisions. This overview provides the reader insights and entry points into learning about two-stage zero-sum games (e.g., minimax or maximin) in which one agent has complete knowledge of the other’s actions. The difference between interdiction and robust optimization models is described, with a focus on steering the reader to relevant and contemporary research in the field.