Optimal control problems via exact penalty functions (Q1387647)

The paper deals with the problem how some constrained optimization problems (like, for instance, control problems) can be reduced to unconstrained ones by the method called exact penalization. This method needs that a suitable function (nonsmooth functional in general) describing the constrained set in the form of equality satisfies some conditions on its directional derivatives. Special attention is paid to the case of optimal control problems for systems described by parametrized ordinary differential equations and a cost functional in integral form. Obtained results permit to replace the minimization of the cost functional (even smooth) on the solution set to the ODE by an unconstrained optimization problem (with the use of Lagrange multiplier) and thus to apply numerical algorithms for unconstrained (but nonsmooth) optimization.