Minimax optimal control of discrete-time uncertain systems with structured uncertainty (Q674769)

The authors present a discrete-time counterpart of their results given in [\textit{A. V. Savkin} and \textit{I. R. Petersen}, J. Optimization Theory Appl. 88, No. 2, 321-337 (1996; Zbl 0853.93065)]. A guaranteed cost control problem (see e.g. [\textit{S. S. L. Chang} and \textit{T. K. C. Peng}, IEEE Trans. Autom. Control 17, 474-483 (1972; Zbl 0259.93018)]) for linear discrete-time systems with structured uncertainties bounded by a sum quadratic constraint is considered. The model of uncertainty allows for a large class of nonlinear time-varying structured uncertainties and provides a measure of the size of initial conditions on the uncertainty dynamics. The existence of the linear state feedback guaranteed controller is guaranteed if there exists a minimal positive definite solution to a parameter dependent Riccati equation. Moreover, the authors prove that the obtained controller is a minimax optimal controller for a specific value of the initial conditon assuming that such a controller does exist. The results obtained for the standard quadratic performance index are then extended to a cost function with cross product terms of control and states. The results are illustrated by a numerical example.