\(H^{\infty}\) robust stabilizability (Q1100173)

A single-input single-output [SISO] linear time-invariant (nondistributed parameter) system can be taken to be given by a proper rational function P(s); multiplication by P(s) represents the input-output operator in the frequency domain. If P(s) is unstable (i.e. P has poles in the right half plane), a goal of control theory is to build another plant C(s), called a compensator, such that the closed-loop system which uses C(s) as a feedback is stable. In practice the plane P(s) is not known precisely; this leads to the robust stabilization problem, namely, to find a compensator which stabilizes every plant P(s) smultaneously in some band of uncertainty around the nominal plant \(P_ 0(s)\). The problem of robust stabilizability was posed and solved by \textit{H. Kimura} [IEEE Trans. Autom. Control AC-29, 788-793 (1984; Zbl 0573.93023)] in 1984 in some particular cases by using Nevanlinna-Pick interpolation techniques. A difficulty of the solution is that there is no guarantee that the stabilizing compensator is proper. Here the authors introduce an approximation procedure which enables them to find proper stabilizing compensators and to estimate the degrees of such compensators.