Reduction of robust stabilization problems to standard \(H^ \infty\) problems for classes of systems with structured uncertainty (Q1923043): Difference between revisions

The paper presents some necessary and sufficient conditions of robust stabilizability for a family of linear time-invariant systems given by transfer function models with structured uncertainty, using linear time-invariant controllers. The line of reasoning goes through reduction of the robust stabilization problem to a standard \(H^\infty\) problem by constructing a nonconservative weighting function and introducing a technical concept of polynomial realizability conditioning this construction. The class of systems for which the relevant \(H^\infty\) problem is equivalent to the considered robust stabilizability includes systems with structured uncertainty both in the denominator and in the numerator but appearing separately. As examples of the first category, the authors present linear time-invariant systems with disc-bounded complex uncertainty entering in an affine way into coefficients of the denominator, systems with real parametric uncertainty and systems with several uncertain real poles. For real uncertainties, the uncertain polynomials should be balanced and the maximal uncertainty magnitude should be realizable via a stable polynomial. It is worth noting that the robust stabilization problem for the system with uncertain denominator is replaced by a standard \(H^\infty\) sensitivity minimization problem while an \(H^\infty\) complementary sensitivity problem is associated with systems containing numerator uncertainty. The authors show also that the applicability of their method depends on the choice of the nominal model for a given plant.