ABSTRACT
This chapter presents the main ideas, philosophy and techniques constituting the identification of state-space models, especially subspace identification methods. Linear algebra-based projection and decomposition algorithms for optimal (Kalman) estimation of states are discussed in detail. A review of the ubiquitous Kalman filter is briefly presented. The chapter also discusses popular specialized algorithms such as N4SID, MOESP and CVA. Subspace methods are equipped with a semi-automated technique for order determination, which is also delineated. This chapter also introduces the ideas of grey-box identification in the context of developing structured state-space models. MATLAB-based examples are presented to illustrate the ideas at each stage of the developments.
