Successive Identification of Volterra Type Nonlinear Multi-Input-Output System and Its Application to Parameter Estimation

Abstract

A simpler practical identification method of nonlinear system represented by discrete Volterra functional series is proposed. Discrete Volterra kernels, divided into group corresponding to each sampling time, are determined from a set of simultaneous linear equations. The sets of simultaneous linear equations are solved one by one, applying the solution of each step to solve the next set of equations having a larger number of unknowns. Thus, each identification procedure reduces the size of matrix to be dealt with and assures an accurate system identification with the combination of least squares estimation method in the presence of noise on output signals. Furthermore, as its application, a parameter estimation method using a part of the solution of linear equations is proposed.