Input Parameter Uncertainty Quantification with Robust Simulation and Ranking and Selection

In this paper, we consider an input parameter uncertainty quantification problem, where the true distributions of the input random factors that drive a simulation model are unknown but belong to some parametric families and we aim to find the worst-case expected performance of the simulation model. We propose a framework to calculate the upper and lower bounds of the worst-case expected performance by adopting robust simulation and ranking and selection methods, respectively. The confidence intervals of the upper and lower bounds are further constructed. The framework can address various parametric families in the input modeling and is flexible in the choice of ranking and selection procedure. Numerical experiments show the effectiveness of the framework.