Black-box identification of multivariable transfer functions—asymptotic properties and optimal input design

Abstract

Abstract The problem of estimating the transfer function of a multivariable linear stochastic system is considered. The transfer function is parametrized as a black box and no order is chosen a priori. This means that the model order may increase to infinity as the number of observed data tends to infinity. The asymptotic covariance of the transfer function estimate is calculated and is found to be independent of the noise model when the model order increases to infinity. The derived expressions are also used to determine optimal ' unprejudiced ' input signals for the identification experiment.