System identification with disturbances

In system identification, the purity of the input and output signals determines the accuracy of the parameter estimates. In many situations the input and output signals are tainted with unmeasurable disturbances other than measurement noise usually considered in system identification. Unlike noise, disturbances can be estimated and corrective measures can be applied during identification. If identification is performed with disturbance effects included in the data signals, the model found will not represent the true system. Therefore, any disturbances must be accounted for and dealt with before or during the identification process. A survey of the various methods proposed in literature for dealing with this problem has been compiled and examined in this paper