An Introduction of a CCA Weighting Matrix to a Closed-Loop Subspace Identification Method

This paper introduces a CCA (canonical correlation analysis) weighting matrix to an estimation method of the innovations model previously proposed by the authors. A numerical simulation illustrates that the CCA weighting reduces the covariance of the estimate and that the proposed method gives similar or better performance compared to Closed-Loop MOESP and PBSID. Especially, the design parameter called "past horizon" in the proposed method can be set small compared to Closed-Loop MOESP and PBSID. It is also analyzed how the bias of the estimate is reduced. Copyright (C) 2021 The Authors.