Analysis of the asymptotic properties of the MOESP type of subspace algorithms (Q1975543)

Subspace algorithms are used for the estimation of linear, time-invariant, finite-dimensional, discrete-time, state-space systems. The paper uses the MOESP type of subspace algorithms. They are based on the geometric structure of covariance matrices and exploit the properties of the state vector extensively. The asymptotic properties of the algorithms are examined. The main results concern consistency and asymptotic normality for the estimates of the system matrices under suitable assumptions. Numerical examples are given, and confidence regions for estimates of different system related quantities can be computed using this asymptotic theory.