Optimality conditions for nonsmooth multiobjective bilevel optimization problems (Q2177790)

The paper is concerned with the finite-dimensional optimistic bilevel vector optimization problem \[ \begin{aligned} &\text{v-}\!\min_{x,y}\{(F_1(x,y),\dots,F_p(x,y) \mid y \in S(x),\, G_k(x) \leq 0,\,1\leq k \leq l\},\\ &S(x) = \text{v-}\!\arg\min_y \{f_1(x,y),\dots,f_q(x,y) \mid g_j(x,y)\leq 0,\, 1\leq j\leq r\}, \end{aligned} \] where \(S(x)\) denotes the set of (weak) Pareto-optimal points of the lower level problem, which is assumed to be convex for each \(x\). All functions are assumed to be merely locally Lipschitz continuous. Following the approach of \textit{S. Dempe} and \textit{A. B. Zemkoho} [SIAM J. Optim. 24, No. 4, 1639--1669 (2014; Zbl 1327.90377)] for bilevel scalar optimization problems, the problem is replaced by its KKT relaxation, where the solution set \(S(x)\) is replaced by the set of KKT points, i.e., points which satisfy first-order necessary optimality conditions in terms of limiting subdifferentials of the lower-level problem. Under a constraint qualification, which in essence requires ``no non-zero abnormal multipliers'', first-order necessary optimality conditions of Fritz John and KKT-type for the bilevel problem in terms of limiting coderivatives of the relaxed lower-level KKT set function.