On robustness in nonconvex optimization with application to defense planning

We estimate the increase in minimum value for a decision that is robust to parameter perturbations as compared of nominal nonconvex problem. The estimates rely on expressions subgradients and local Lipschitz moduli min-value functions require only solution Across 54 mixed-integer optimization models, median error estimating 12%. results inform analysts about possibility obtaining cost-effective, parameter-robust decisions.