Direct performance optimization using laguerre models

Abstract

This paper presents a tuning method based on direct optimization. A model is first obtained by injecting a sum of sinusoids as an excitation signal; an off-line method is used to obtain the coefficients of a truncated Laguerre series model. Given this model, the gradient and Hessian with respect to the controller parameters are computed off-line, using only error-signal measurements. Parameters are adjusted through Newton-Raphson iterations.