Robust adaptive identification of linear time-varying systems with modified least-squares algorithm.

A modified least-squares algorithm with forgetting factor is proposed and applied to linear time-varying systems. The bound on the Euclidean norm of the estimation error is derived mathematically. When the bounded regressor of the linear time-varying system satisfies persistent excitation condition, this new algorithm can attenuate the influences of bounded disturbances and the change ratio of timevarying parameters effectively through extra term that is added to the covariance's update law. Contrary to the earlier proposed estimation laws, this algorithm does not require any prior knowledge of upper bounds on the parameters and the disturbances. Simulation results demonstrate the effectiveness of the identification law. [ABSTRACT FROM AUTHOR]