Project Grant 2246281

This National Science Foundation Project Grant award of $459,173 supports research advancing the theory and practice of robust simulation analysis under input model risk. Funded through the Engineering program (CFDA 47.041), the awardee, Georgia Tech Research Corporation, will develop a robust decision-making framework to quantify, reduce, and ensure robust decisions when probability distribution functions driving randomness in simulation models are estimated based on available data, introducing "input model risk." Over the four-year period from July 2022 to August 2026, the awardee will study methods to balance practical user input on acceptable suboptimality with statistical robustness to errors in simulation models. A minimum-cost simulation experiment design and Gaussian process metamodel will be created to predict simulation outputs accounting for both parametric and nonparametric input models. This framework aims to reduce conservatism while achieving user-desired robustness levels, defining and finding a "practically robust" optimal solution through efficient simulation-guided optimization algorithms. The project also provides guidance to reduce input model risk by optimizing data collection plans for stronger performance guarantees on practical robustness.