A matrix Euclidean algorithm and matrix continued fraction expansions (Q5896423): Difference between revisions

In previous papers by \textit{R. E. Kalman} [Acta Polytech. Scand., Math. Comput. Sci. Ser. 31, 9-32 (1979; Zbl 0424.93020)], \textit{W. B. Gragg} and \textit{A. Lindquist} [Linear Algebra Appl. 50, 277-319 (1983; Zbl 0519.93024)] the Euclidean algorithm, to which \(V_{\max}\) is directly related, was used for producing a nested sequence of partial realizations and obtaining a continued fraction representation of a strictly proper transfer function. Kalman derives, as a by product, a characterization of the maximal (A,B)-invariant subspace in Ker C for a minimal realization (A,B,C) which is associated with the continued fraction expansion. This paper reverses the reasoning and starts with a simple idea drawn from the Morse-Wonham geometric control theory, namely the knowledge that \(V^*_{\ker C}\) is related to maximal McMillan degree reduction by state feedback. In fact, a minimal system is feedback irreducible if and only if \(V^*_{\ker C}=\{0\}.\) This property is used for the construction of a continued fraction representation, or for a version of the Euclidean algorithm for the computation of \(V^*_{\ker C}.\) A basic feature of the paper is that it introduces the dual concept of reduction by output injection and uses it for deriving most of the results in a simple way. A recursive characterization of \(V^*({\mathcal B})\) is also derived.