Auxiliary Model Maximum Likelihood Moving‐Data‐Window Generalized Extended Gradient‐Based Iterative Algorithm for Multivariable Autoregressive Output‐Error Autoregressive Moving‐Average Systems

ABSTRACT This article considers the parameter estimation problems of multivariable autoregressive output‐error autoregressive moving‐average systems. To alleviate the identification difficulty of systems, we decompose the multivariable autoregressive output‐error autoregressive moving‐average system into several subsystems. To reduce the impact of colored noises on parameter estimation accuracy, we propose an auxiliary model maximum likelihood generalized extended gradient‐based iterative algorithm using the auxiliary model idea and the maximum likelihood principle. In addition, we use the multi‐innovation identification theory to introduce a moving data window and propose an auxiliary model maximum likelihood moving‐data‐window generalized extended gradient‐based iterative algorithm. These two proposed algorithms produce more accurate parameter estimates than the existing auxiliary model generalized extended gradient‐based iterative algorithm. Simulation examples illustrate the effectiveness of the proposed algorithms.