Simulation optimization when facing input uncertainty

Stochastic simulation is driven by the input model, which is a collection of distributions that model the randomness in the system. The input model is often constructed from data, and hence input uncertainty arises due to the finiteness of data. Simulation optimization has been mostly studied under the assumption of a known input model, without accounting for input uncertainty. We propose a new framework to study simulation optimization under input uncertainty, with the goal to balance the trade-off between optimizing under the estimated input model and hedging against the risk brought by input uncertainty. A simple numerical example illustrates different formulations under the new framework, compared with the usual formulation for simulation optimization.