System-theoretic and algebraic aspects of the rings of stable and proper stable rational functions (Q1061692)

This paper represents a seminal contribution to the current interest in the use of matrix fractional representation techniques for linear systems. In particular the work of \textit{P. A. Fuhrmann} [J. Franklin Inst. 301, 521-540 (1976; Zbl 0332.93001) and Int. J. Control 25, 5-10 (1977; Zbl 0357.93009)] in relating polynomial matrix methods to state- space techniques is extended to fractional representations over rings of stable and proper stable rational functions. Three areas are investigated in depth: (i) Several unique decompositions for the field of rational functions are presented: \(\Lambda =\Lambda_+\oplus_ K\Lambda_-\) and \(\Lambda =\Lambda_ s\oplus_ K\Lambda_ u\) where \(\Lambda\) is the field of rational functions in the indeterminate z with coefficients in a fixed field K, \(\Lambda_+\) is the ring of polynomials in z with coefficients in field K, \(\Lambda_-\) is the ring of strictly proper rational functions, \(\Lambda_ s\) is the ring of stable rational functions and \(\Lambda_ u\) is the ring of strictly proper completely unstable rational functions. The discussion then moves on to the characterization of two projections \({\hat \Pi}\): \(\Lambda\) \(\to \Lambda_ u\), and \(\Pi_{\nu}: \Lambda_ s\to \Lambda_ s\) and the subspace \(X_{\psi}{\underset {=} \Delta} im \Pi_{\nu}\) and the null space ker \(\Pi\) \({}_{\nu}\). This is followed by the matrix version \(\Pi_ v\), \(X_ v{\underset {=} \Delta} im \Pi_ v\), and ker \(\Pi\) \({}_ v\). The ring structure is completed by an analysis of the ring of proper stable rational functions denoted \(\Lambda_{ps}\). The important decomposition: \(\Lambda =\Lambda_{ps}\oplus_ K\Lambda_{au}\) where \(\Lambda_{au}{\underset {=} \Delta}z\Lambda_+\oplus_ K\Lambda_ u=\{z\gamma +\beta^{-1}\alpha:\quad \gamma \in \Lambda_+,\quad \beta^{-1}\alpha \in \Lambda_ u\}\) is obtained. Scalar and matrix characterisations for the projections \(\Pi\) : \(\Lambda\) \(\to \Lambda_{au}\), \(\Pi_ q\), subspace \(X_ q{\underset {=} \Delta} im \Pi_ q\), ker \(\Pi\) \({}_ q\) projections \(\Pi_ Q\), subspace \(X_ A{\underset {=} \Delta} im \Pi_ Q\) and ker \(\Pi\) \({}_ Q\) close the section. (ii) The second area covered is the problem of finding a state space realization for fractional representations over the rings \(\Lambda_ s\), and \(\Lambda_{ps}\). The module-theoretic approach of \textit{R. E. Kalman}, \textit{P. L. Falb} and \textit{M. A. Arbib} [Topics in mathematical system theory (1969; Zbl 0231.49001), Chapter 10] is utilised and the relationship with the state-space realizations for polynomial based transfer functions explored. The reachability and observability properties of the state space realization is related to several coprimeness conditions. Fractional representations over both \(\Lambda_ s\) and \(\Lambda_{ps}\) are analysed. (iii) Existence conditions and solution procedures are given for two linear equations important in linear systems control problems. The unilateral equation \(VX+WY=T\) where V, W and T are in \(\Lambda_ s^{r\times r}\), \(\Lambda_ s^{r\times m}\) and \(\Lambda_ s^{r\times r}\) respectively, with X and Y (in \(\Lambda_ s^{r\times r}\), \(\Lambda_ s^{m\times r}\) respectively) as the required unknown matrices, is pursued first. This is followed by a similar presentation for the bilateral equation \(VX+YW=T\) where \(V\in \Lambda_ s^{r\times r}\), V nonsingular, \(W\in \Lambda_ s^{m\times r}\), \(T\in \Lambda_ s^{r\times r}\) and X,Y are the unknown matrices (over \(\Lambda_ s)\). The realisation procedures figure prominantly in the theorems whose constructive proofs also provide solution procedures. The results for the bilateral equation are illustrated by a simple example.