State-space realizations of MIMO 2D discrete linear systems Elementary operation and variable inversion approach (Q4485359)

An elementary operations algorithm (EOA), introduced by the author in some previous papers, is combined with variable inversion in order to obtain nonsingular realizations for 2D transfer functions. The basic principles of EOA are presented, and some new comments are provided in order to clarify the methodology. The starting point is the association of a 3D polynomial matrix to a rational 2D transfer matrix by introducing a new complex variable. It is shown that augmentations and elementary (row and column) operations, which are basic for EOA, performed on polynomial matrices leave a given transfer matrix invariant. This result allows the relaxation of some requirements present in a previous paper of the author and increases the possibility of deriving equivalent realizations. It is emphasized that the possibility to obtain singular realizations is a considerable limitation of the method. The aim of the paper is to improve the procedure in order to obtain nonsingular realizations for much wider classes of principal transfer matrices. By using the inversion of variables, it is shown that a singular realization can be transformed in a standard (nonsingular) form if some submatrix is non-singular. This sufficient condition is characterized by a property of the transfer matrix. A method is also provided to reduce the dimension of the above realizations. The possibility of exploiting symbolic computing techniques is emphasized by using Maple V notations. Some examples are given to clarify the methods and to emphasize their advantages.