A smoothing Newton method for mathematical programs constrained by parameterized quasi-variational inequalities (Q657283)

The authors consider the mathematical programming problem governed by parametrized generalized equations \[ \text{minimize \(\varphi(x,y)\) subject to }0 \in g(x,y) + \mathcal{N}_{\Gamma (x,y)}(y),\tag{1} \] where \(\varphi: \mathbb R^n \times \mathbb R^m \rightarrow \mathbb R, g: \mathbb R^n \times \mathbb R^m \rightarrow \mathbb R^m\) are twice continuously differentiable mappings, \(\mathcal{N}_C (z)\) denotes the (limiting) normal cone of a set \(C \subset \mathbb R^m\) at \(z \in C\) and \( \Gamma: \mathbb R^n \times \mathbb R^m \rightrightarrows \mathbb R^m\) is a set - valued mapping which admits the representation, \[ \Gamma (x,y) = \{z \in \mathbb R^m\mid q (x,y,z) \in \mathbb R^s_- \},\tag{2} \] with \(q: \mathbb R^n \times \mathbb R^m \times \mathbb R^m \rightarrow \mathbb R^S\) (a twice continuously differentiable vector function). The constraint in problem (1) (if \(\Gamma\) is a convex-valued mapping) becomes the parametrized quasi-variational inequality (QVI): find \(y \in \Gamma (x,y) \subset \mathbb R^m\) such that \[ \langle g (x,y), z-y \rangle \geq 0,\quad \text{for all } z \in \Gamma (x,y),\tag{3} \] and problem (1) is reduced to the mathematical program governed by quasi-variational inequality constraints. The authors consider the mathematical programming problem with equilibrium constraints expressed in (1), where \(\Gamma\) is defined in (2) and is a convex-valued mapping. Second order sufficient conditions for problem (1) which are helpful to prove the quadratic convergence of the smoothing Newton method are given. The necessary conditions for problem (1) in terms of the coderivates as a system of nonsmooth equations under a linear independence constraint qualification and the strict slackness condition are expressed. This nonsmooth equation reformulation is new and the key to construct a smoothing Newton method for finding a stationary point (in the sense of Mordukhovich). For the purpose of convergence analysis, the authors introduce a set of second order sufficient conditions for problem (1) which are sufficient conditions for the second order growth. The strong BD-regularity of the nonsmooth system is demonstrated under the proposed second order sufficient conditions, linear independence constraint qualification and the strict slackness condition. Main result: The paper is devoted to deriving necessary optimality conditions, for mathematical programs with parametrized QVI constraints, in the form of non-smooth equations, and proposing a set of second order sufficient conditions. This paper develops a smoothing Newton methods for solving the nonsmooth equations for a stationary point. Necessary conditions for a local minimizer of problem (1) are characterized by the system of nonsmooth equations, \(F(x,y,d,v,u) = 0\), under some mild conditions, where \(F(x,y,d,v,u) = \langle \nabla \varphi (x,y) + (\tau_{x,y} L(x,y,d))^T v + \tau r (x,y))^T u, L (x,y,d), \min (d_1, - r_1 (x,y)) \dots \min (d_s, - r_s (x,y)) \min (0, - d_1)(p(x,y)v)_1 + \min (0, r_1(x,y) u_1 \dots \min (0, -ds)(p(x,y)v)_s + \min (0, r_s(x,y))u_s \rangle^T\). The authors propose the precise proofs of the theorems. Numerical results for solving seven test problems are also presented.