A computational procedure for suboptimal robust controls (Q1200602): Difference between revisions

This paper is devoted to the study of the sensitivity of the numerical solution of a general nonlinear optimal control problem with respect to both: the parameters \(\sigma\) of discretization of the problem and an uncertain parameter \(\alpha\) which influences the right-hand-side of the state equation, the initial state and the data of the Bolza-type performance index \(J\). The main result consists of a formula for the differential of the performance index \(G\) with respect to \(\sigma\), where \(G=\widetilde J+c\|\text{grad}_ \alpha \widetilde J\|^ 2\), \(c\) is a constant, and \(\widetilde J\) is the value of the original performance index \(J\) in the class of piecewise-constant approximations of admissible controls, defined by \(\sigma\). Two numerical examples are presented.