The Identification of Multivariable Systems Based on a New Two-Stage Hierarchical Least-Squares Iterative Algorithm

Abstract

In this paper, a multivariable system with moving average noises can be expressed as a multivariable CARMA model. To solve the difficulty of the information matrix including unknown noise terms, the noise is estimated by fitting a high order CAR model by using the hierarchical least-squares iterative (HLSI) method. Based on the obtained noise, a modified HLSI method for identification of multivariable systems is proposed. It consists of two-stage HLSI method which can be on line applied to the identification of multivariable systems, and has fast convergence rate. Some simulations illustrate the effective of the proposed approach.