Convexity and closedness in stable robust duality (Q2414114)

The linear disturbed robust primal problem $(\mathbf{RP})_{x^*}$: $\sup_{u\in U} F_u(x,0_u)-\langle x^*,x\rangle \rightarrow \inf_{x\in X}$ with $0_u$ as zero element of $Y_u$ and the corresponding robust dual problem $(\mathbf{ODP})_{x^*}$: $-F^*_u(x^*,y^*_u)\rightarrow \sup_{\{(u,y^*_u)\mid u\in U, y^*_u\in Y^*_u\}}$ are considered where $X,Y_u, u\in U$ are locally convex Hausdorff topological spaces, $U$ is a uncertainty parameter set and $X^*,Y^*_u$ are the spaces of corresponding continuous linear functionals w.r.t. the weak$^*$ topology. The investigations on stable robust duality started in [3, \textit{N. Dinh}, \textit{M. A. Goberna}, \textit{M. A. López} and \textit{M. Volle}, ``Characterizations of robust and stable duality for linearly perturbed uncertain optimization problems'', Preprint, \url{arxiv:1803.04673}] are continued to the special case whenever the negative optimal value function $x^*\mapsto -q(x^*)$ of the above dual problems is convex or convex and lower semi-continuous. Several stable (strong) robust duality results are presented which use the stable robust duality Theorem 6.1 and Theorem 7.1 of [3]. If $U$ is a singleton, i.e. $Y_u\equiv Y, Y\sqsubseteq X$, and $F(x,y):=f(x+y)+g(y)$ then it follows from their duality results an extension of [\textit{J. M. Borwein} et al., J. Nonlinear Convex Anal. 15, No. 1, 167--190 (2014; Zbl 1301.49035), Theorem 3.2] and as consequence that $(f+g)^*=f^*\square g^*$ is valid either under the Bertsekas constraint qualification or weak$^*$ closedness of some union of sums of $\varepsilon$-subdifferentials of $f$ and $g$. For understanding the ideas used in the paper it is very convenient first download and study [3].