A bias-correction method for indirect identification of closed-loop systems (Q1899582)

The paper deals with indirect identification of closed-loop systems in the presence of coloured noise. A least-squares-like method, with an incorporated bias correction procedure, is developed for consistent estimation of the closed-loop system and open-loop process. The proposed method exploits in a novel way the available information regarding the regulator (feedback law) in the first step of the indirect identification procedure, yielding a simple and efficient procedure for computing the bias of the conventional LS estimate of the closed-loop parameters. It is shown that the bias-eliminated LS estimate, obtained after removing the bias, besides consistent convergence is robust against different kinds of noise models, exhibits a low sensitivity to SNR changes, and retains all computational merits of the standard LS method. These advantages are illustrated in the paper by means of a simulation example.