A class of exact penalty functions and penalty algorithms for nonsmooth constrained optimization problems (Q2307749)

For nonlinear optimization problems with equality, inequality and bound constraints, the authors consider a family of penalty functions and prove that, under suitable assumptions, which include a so-called weakly generalized Mangasarian-Fromovitz constraint qualification, when the penalty parameter is large enough every local optimal solution (stationary point) of the penalty problem yields a local optimal solution (respectively, stationary point) of the original problem, and every local optimal solution of the original problem can be obtained in such a way. They present an associated globally convergent exact smoothing penalty method, which allows for feasibility verification, and report some numerical results.