Subspace Identification of 2-D CRSD Roesser Models With Deterministic-Stochastic Inputs: A State Computation Approach

In this brief, we present a subspace system identification framework for 2-D separable-in-denominator systems with deterministic-stochastic inputs in the Roesser form. The advantage of the proposed framework is that it is based on the computation of state matrices, as opposed to current algorithms that compute the system parameter matrices from Markov parameters and the observability matrix. As such, it does not require solving specialized Toeplitz or Hankel systems of equations while computing the system parameter matrices. In addition, the problem is broken down into two simple oblique projection computations-one in the horizontal direction and one in the vertical direction. Within this framework, Numerical algorithms for Subspace State Space System IDentification (N4SID), Past-Output Multivariable Output-Error State-sPace (PO-MOESP), and Canonical Variate Analysis (CVA) type algorithms are obtained. Simulation results show that the algorithms are accurate and provide new alternatives for modeling and identifying 2-D causal, recursive, and separable-in-denominator Roesser models.