Gain-scheduling procedures for unstable controllers: a behavioral approach

A behavioral approach to the design of gain-scheduled controllers is presented. Plant models are considered to be black-box models obtained at finitely many operating points, and it is assumed that no analytical nor empirical relations are known between physical parameters and the plant models. In this situation, it is suggested to interpolate in intermediate operating points, simply by interpolating in signal space in the behavioral formalism. Two controller architectures are suggested that lead to global stability for systems that remain close to the interpolated values.