Iterative Weighted Least Squares Identification and Psrm Control Design

This paper presents an iterative refinement scheme of successive closed-loop identifications and control law designs that aim to improve iteratively the achieved closed-loop performances. The key to this approach is to account for the evaluated modelling error in the control design and to let the closed-loop controller requirements determine the identification criteria. The first is achieved by the partial state reference model (PSRM) control criterion with frequency weighting filters that reflect the mismatch between the actual and the designed closed loop systems. The PSRM approach has been motivated by its suitable tracking capability. The least-squares identification is performed on closed-loop data with a filter, which improves the model accuracy at those frequencies where the robust performance dictates that a better model is needed. The effectiveness of the proposed iterative design strategy is illustrated by an example.