Robust Analysis in Stochastic Simulation: Computation and Performance Guarantees

Decision making using simulation analysis faces the risk of being adversely affected by insufficiently calibrated input models. This risk is especially significant when only limited historical observations of the real-world process are available, or when unexpected risk factors are suspected and require elicitation of expert opinion. In “Robust Analysis in Stochastic Simulation: Computation and Performance Guarantees”, Ghosh and Lam propose a new methodology that solves robust optimization formulations over generic simulation models to compute bounds on the worst possible performance values. Precise performance guarantees are derived for the proposed methods, and these are further illustrated with key examples from stochastic service and queueing systems.