Зведення задачі мінімаксного керування лінійними нестаціонарними системами до – робастного шляхом динамічної гри

Abstract

The problem of synthesis of minimax control for the dynamic, described by the linear system of differential equations (taking into account the state, controls, perturbations and initial conditions, with the given equation of observation inclusive) of objects functioning in accordance with the integral-quadratic quality criterion in uncertainty is solved in the work. External perturbations, errors, and initial conditions were assumed to belong to a number of uncertainties. The task of finding optimal control in the form of a feedback object that minimizes the performance criterion is presented in the form of a minimum maximal uncertainty control problem. In the absence of ready-made solution paths, this problem is reduced to a -control problem under the most unfavorable disturbances, and in addition to a dynamic game problem with zero sum and a certain price for the game, and a strategy for solving it is proposed that offers a way to new results. The problem of finding the optimal control and the initial state that maximize the quality criterion is considered in the framework of the optimization problem solved by the Lagrange multiplier method after introducing the auxiliary scalar function, the Hamiltonian. It is shown that to find the maximum value of the criterion, either the necessary condition of the extremum of the first kind can be used, which depends on the ratio of the first variation of the criterion and the first variations of the control vectors and the initial state, or also the necessary condition of the extremum of the second kind, which depends on the sign of the second variation. For the first and second variations, formulas are given that can be used for calculations. It is suggested to solve the control search problem in two steps: search for an intermediate solution at fixed values of control vectors and errors, and then search for final optimal control. Consideration is also given to solving -optimal control for infinite control time with respect to the signal from the compensator output, as well as solving the corresponding Riccati matrix algebraic equations.