Abstract
A robust recursive method for parameter estimation of linear lumped systems via Hermite polynomials has been derived. The main emphasis is on the algorithm developed to estimate the coefficients of the Hermite series expansion of the process input-output signals using Huber's min-max principle, when the measurements are corrupted with contaminated normal noise. The properly chosen performance criterion ensures immunity to the occurrence of outliers in the data. The robust estimate of the system parameters is obtained by the use of the robust estimate of the expansion coefficients. A numerical example is presented to illustrate the algorithm.