Recursive maximum likelihood identification method for a multivariable controlled autoregressive moving average system

Abstract

The identification model of the multivariable controlled autoregressive moving average (CARMA)-like system contains a parameter matrix and a parameter vector, and cannot be performed by a standard least squares method. In order to solve this identification problem, we transform the CARMA-like system into |$m$| regression identification models (⁠|$m$| is the number of outputs) each of which has only a parameter vector, and use a recursive maximum likelihood identification algorithm to estimate the parameter vectors of these submodels. The proposed algorithm is simple and effective, and has high estimation accuracy.