Analysis Of Subspace Identification Methods Based On The Estimation Of The System Matrices

The subspace identification methods or 4SID methods are known for their ability to construct mathematical models in state space from the input-output data of an LTI system. In these algorithms, the calculation of extended observability matrix is the step that receives the greatest attention, and generally, the performance of the algorithms is evaluated based on this matrix parameters. In contrast, the step of calculating the matrices of the system often does not receive the same attention, but their influence on the quality of the resulting model may be remarkable. In this paper we describe the principal techniques for calculating the system matrices and analyze their effect, via simulation, on the resulting mathematical model.