Metric inequality conditions on sets and consequences in optimization (Q2046537)

If we consider a constrained optimization problem, it is known that to obtain necessary optimality conditions some constraint qualification conditions are necessary. If we add to the system of constraints a new constraint, the problem can change and if the initial system satisfies a constraint qualification, the new system can fail to do so. In this work, the authors are trying to give a condition that links the old system of constraints and a new one with a more constraint, in such a way that the optimality conditions apply for this new problem. First, they study a smooth optimization problem and arrive to a metric condition which is equivalent to another well-known metric inequality condition studied under various names (metric inequality, linear regularity, metric regularity, linear coherence, or subtransversality). They analyze this metric condition related to other well-known constraints qualifications. An important characteristic of the metric condition is that it can be employed in nonsmooth frameworks. Then, the obtained results are applied to prove necessary optimality conditions to directional minima. Finally, the metric condition is applied to penalize scalar nonsmooth optimization problems with multiple constraints, and also to derive necessary optimality conditions in terms of limiting generalized differentials.