Canonical state representations and hilbert functions of multidimensional systems

A basic and substantial theorem of one-dimensional systems theory, due to R. Kalman, says that an arbitrary input/output behavior with proper transfer matrix admits an observable state representation which, in particular, is a realization of the transfer matrix. The state equations have the characteristic property that any local, better temporal, state at time zero and any input give rise to a unique global state or trajectory of the system or, in other terms, that the global state is the unique solution of a suitable Cauchy problem. With an adaption of this state property to the multidimensional situation or rather its algebraic counter-part we prove that any behavior governed by a linear system of partial differential or difference equations with constant coefficients is isomorphic to a canonical state behavior which is constructed by means of Grobner bases. In contrast to the one- dimensional situation, to J.C. Willems' multidimensional state space models and and to J.F. Pommaret's modified Spencer form the canonical state behavior is not necessarily a first order system. Further first order models are due E. Zerz. As a by-product of the state space construction we derive a new variant of the algorithms for the computation of the Hilbert function of finitely generated polynomial modules or behaviors. J.F. Pommaret, J. Wood and P. Rocha discussed the Hilbert polynomial in the systems theoretic context. The theorems of this paper are constructive and have been implemented in MAPLE in the two-dimensional case and demonstrated in a simple, but instructive example. A two-page example also gives the complete proof of Kalman's one-dimensional theorem mentioned above. We believe that for this standard case the algorithms of the present paper compare well with their various competitors from the literature.