On the quality of recursively identified FIR models

The author considers recursive identification of time-varying systems having finite impulse response, focusing on the tradeoff between tracking capability and disturbance rejection. Approximate, but simple and explicit, frequency-domain expressions for the model quality are derived for three different identification algorithms. The results, derived under the assumption of slow adaptation, slow system variation, and high model order, are extensions of the results presented by Gunnarsson and Ljung (see ibid., vol.37, p.1072, 1989) to the case where the system output is affected by correlated disturbances. >