Control-Relevant Uncertainty Modelling Directed Towards Performance Robustness

Abstract In line with recently introduced methods for iterative identification and control design, a method is proposed to identify uncertainty models from data with an uncertainty structure that is motivated by the (closed loop) control performance cost function. This allows the assessment of achieved plant performance on the basis of measured time series, as well as the evaluation of robust stability and robust performance for a newly designed controller prior to implementation. For the specific control performance measure considered, ( H ∞ -norm on a closed-loop transfer matrix), this naturally leads to the identification of upper bounds on either coprime factor model uncertainty or uncertainty on the (dual) Youla-parameter. Theoretical results are supported by experimental results of an application to the radial control loop in a compact disc servo mechanism.