What are Subspace Methods?
They are a class of methods for the identification of linear state space models directly from the experimental input/output data. SIM are an alternative to the well known prediction error identification of ARX and ARMAX models. However, they are based on a very different approach of the geometric projections and linear algebra. The basic ideas of these methods were developed about 20 years ago and they are quite well accepted in the control engineering community, however the applications of these methods in the systems identification are still rather exceptional, which is mainly due to their more complex theoretical background. The name ‘Subspace’ stands for the fact, that the state space model parameters (matrices A,B,C,D) can be extracted from the row or column subspace of certain matrices, which are obtained only from input/output data.
