Liang J.

This paper extends an existed off-line identification scheme, by incorporating a moving algorithm, so that on-line identification of continuous-time systems can be performed. The algorithm based upon Haar wavelet expansions uses the observed input-output data to estimate unknown system parameters of the nonlinear systems. A recursive identification formula augmented with indirect matrix-inversion schemes is developed. By means of the proposed algorithms, the estimates can be updated step by step without repeatedly computing the matrix inversion. Thus, the computation effort and the computer memory required in the identification process are minimized. The algorithm is validated through numerical and experimental examples. The results indicate the reliability and effectiveness of the proposed approach.