A new method for obtaining on-line unbiased parameter estimates of linear systems in the presence of additive uncorrelated noise is described. The method is based on the "equation error" and is presented in the form of continuous variables. Instead of the conventional approach of least squares, the autocorrelation function of the error is chosen as the cost function to be minimized. It is shown that for comparatively wide band noise, this cost function yields practically unbiased estimates. The solution of the estimate retains the main advantages of the equation error, i.e., uniqueness and often asymptotic stability. On-line identification from system input and output is feasible both in open and closed loop. The method is illustrated by numerical examples.
展开▼
