A Novel Non-Iterative Mimo State Space Model Identification Technique

Abstract The recently proposed family of MOESP, standing for MIMO Output-Error State space model identification, algorithms is extended in this paper. The extension consists of the (efficient) incorporation of instrumental variables in the MOESP algorithmic structure. In particular, the extension is described for the ordinary MOESP scheme. Two types of instrumental variables are discussed: (1) Based on past input measurements and (2) Based on reconstructions of the state sequence of the underlying deterministic system. The latter reconstruction is based on the estimated state space model obtained with the ordinary MOESP scheme. Algorithmically, this extension retains the striking simplicity of the MOESP family of algorithms, namely only a single RQ factorization, a single SVD and the solution of 2 sets of equations is involved. Practically, the usefulness of this extension is demonstrated in the identification of the longitudinal aircraft dynamics using (simulated) flight test data recorded when flying through a vertical gusty wind field.