Characterizing Optimality for a Class of Nonconvex Quadratic Robust Optimization Problems Bilaterally Quadratically Constrained Under Interval Uncertainty

F. Flores-Bazán ; A. Pérez

Journal of convex analysis, Tome 31 (2024) no. 1, pp. 25-38

Résumé

This paper analyzes the following robust optimization problem: \begin{equation*} \smash{\min\Big\{\dfrac{1}{2}x^\top Ax+a^\top x :~\alpha\leq \dfrac{1}{2}x^\top Bx+b^\top x+c\leq\beta,~\forall~(B,b)\in{\mathcal B}_0\Big\}, } \end{equation*} where $\mathcal{B}_0\doteq\{B_1+\mu B_2:\mu\in[\mu_1,\mu_2]\} \times\{b_1+\delta b_2:\delta\in[\delta_1,\delta_2]\}$, with all the matrices involved are real symmetric, $a,b\in\mathbb{R}^n$ and $\alpha,\beta,\delta_1,\delta_2,\mu_1,\mu_2$ are given real numbers. To be more precise, we establish characterizations of the fulfillment of: (i) the robust alternative result; (ii) the robust S-lemma, and (iii) the robust optimality, to the problem above. To that purpose, we apply the convexity result proved by one of the authors valid for nonhomogeneous quadratic functions, instead of the Dines convexity theorem.

Détail

Classification : 90C20, 90C30, 90C26, 90C46
Mots-clés : Nonconvex quadratic programming under uncertainty, robust optimization, S-lemma, global optimality