Identification of Continuous-time Hammerstein Models Using Simultaneous Perturbation Stochastic Approximation

This paper performs an initial study on identification of continuous-time Hammerstein models based on Simultaneous Perturbation Stochastic Approximation (SPSA). While the structure information such as the system order is available for the linear subsystems, the structure of nonlinear subsystem is assumed to be completely unknown. For handling it, a piecewise-linear functions are used as a tool to approximate the unknown nonlinear functions. The SPSA based method is then used to estimate the parameters in both the linear and nonlinear parts based on the given input and output data. A numerical example is given to illustrate that the SPSA based algorithm can give an accurate parameter estimation of the Hammerstein models with high probability through detailed simulation.