Abstract: The stability of a gradient algorithm for timevarying parameter identification is studied in this paper. The stability of the timevarying parameter identification gradient algorithm is analyzed by using the stochastic process boundedness criterion. The sufficient conditions to ensure the stability of the gradient algorithm are demonstrated. It is shown that if the rate of variation of the parameter to be identified is bounded, and the measurement noise is zeromean white noise, under the persistent excitation condition, the identification error given by the properly designed gradient algorithm is bounded in the sence of mean square. This analysis is different from prior works in that it is not necessary to assume that the parameter variation is the zeromean white noise in the proof of the theory.

Key words: parameter identification, gradient algorithm, stochastic process, stability