An uncertainty averaging approach to optimal guaranteed cost control of uncertain systems with structured uncertainty (Q1905927)

The paper under review extends certain previous results of the authors, Petersen and McFarlane, and others, in the time-varying linear-quadratic regulator problem with structured uncertainty. An approach to a feedback design for a robust linear-quadratic system with uncertainty, which uses an upper bound for a sequence of probabilistic functions as the cost functional, was suggested by Petersen and McFarlane in 1992, who called it the guaranteed cost control approach. The authors consider this approach with structured uncertainty. (``Structured'' means that the uncertainty can be broken into a number of independent blocks.) In their formulation, the initial conditions on the uncertainty dynamics have an elementary probabilistic interpretation. The main result of the paper consists in proving necessary and sufficient conditions for the existence of a nonlinear feedback guaranteed cost control. This condition is presented in terms of the existence of a solution to a corresponding Riccati differential equation.