Closed-loop state feedback for optimization of uncertain control systems. II: Multiply closed feedback (Q1287299)

In Part I [Autom. Remote Control 57, No. 7, Pt. 2, 1008--1015 (1996); translation from Avtom. Telemekh. 1996, No. 7, 121--130 (1996; Zbl 0932.93034)], the authors showed that closable feedback greatly enhanced the effectiveness of positional control. Although its exact realization involves considerable expense both at the preliminary stage (before commencement of the control process) and in the course of an actual control process, there does exist a natural approximation of auxiliary operations that admits realization via modern computing techniques. In this paper, the preceding results are extended to multiple contact moments. The realizations are described with the help of microprocessor devices of optimum feedbacks, an intermediate step between open-loop feedbacks and classical closed-loop feedbacks. As is known, in the determined problems of optimum control the latter two types of feedbacks coincide. The distinctions arise in research of the problems of optimum control under the conditions of uncertainty. The construction of closed optimum feedbacks, connected with the solution of the Isaacs-Bellman equation, presents the extraordinarily difficult problem for the tasks with geometrical restrictions on the control. In some sense, the problem of guaranteed optimization, considered in the paper, adjoins to the theory of differential games. Therefore the received results may be interpreted as the approximate solution of a special class of differential games. They can find application for the solution of many applied problems (for example, the stabilization problem) in which the constantly acting unknown disturbances play a considerable role.