Gilbert J. Groenewald

Bounded Real Lemma and structured singular value versus diagonal scaling: the free noncommutative setting

The structured singular value $μ$ was introduced independently by Doyle and Safanov as a tool for analyzing robustness of system stability and performance in the presence of structured uncertainty in the system parameters. While the structured singular value provides a necessary and sufficient criterion for robustness with respect to a structured ball of uncertainty, it is notoriously difficult to actually compute. The method of diagonal (or simply "D") scaling, on the other hand, provides an easily computable upper bound (which we call $\hat μ$) for the structured singular value, but provides an exact evaluation of $μ$ (or even a useful upper bound for $μ$) only in special cases. However it was discovered in the 1990s that a certain enhancement of the uncertainty structure (i.e., letting the uncertainty parameters be freely noncommuting linear operators on an infinite-dimensional separable Hilbert space) resulted in the $D$-scaling procedure leading to an exact evaluation of $μ_{\text{enhanced}}$ ($μ_{\text{enhanced}} = \hat μ$), at least for the tractable special cases which were analyzed in complete detail. On the one hand this enhanced uncertainty has some appeal from the physical point of view: one can allow the uncertainty in the plant parameters to be time-varying, or more generally, one can catch the uncertainty caused by the designer's decision not to model the more complex (e.g. nonlinear) dynamics of the true plant. On the other hand, the precise mathematical formulation of this enhanced uncertainty structure makes contact with developments in the growing theory of analytic functions in freely noncommuting arguments and associated formal power series in freely noncommuting indeterminates. In this article we obtain the $\widetilde μ= \hat μ$ theorem for a more satisfactory general setting.