Subspace Propagator-based Approach for Online State-space System Identification

This paper deals with the problem of linear state-space identification of time-varying dynamic systems. The proposed approach relies on a recursive formulation of the Multi-Input Multi-Output Output Error State-space algorithm and on the so called propagator. In this framework each instantiation of the observation vector is computed through the QR decomposition, which is recursively updated by applying Givens rotations, while the extended observability matrix is updated by using the propagator approach. The effectiveness of the proposed framework is demonstrated in a practical case study represented by a mild nonlinear time varying system.