Abstract-Row reduced representations of behaviors over fields posses a number of useful properties. Perhaps the most important feature is the predictable degree property. This property allows a finite parametrization of the module generated by the rows of the row reduced matrix with prior computable bounds. In this paper we study row-reducedness of representations of behaviors over rings of the form Zpr , where p is a prime number. Using a restricted calculus within Zpr we derive a meaningful and computable notion of row-reducedness.
I. MOTIVATION AND PROBLEM STATEMENTSIn the behavioral theory, the central role is played by the set B of trajectories that characterize a dynamical system Σ, see the textbook [11]. In fact, a dynamical system is defined as a triple Σ = (T, W, B), where T is the time axis, W is the signal alphabet, and where B, the behavior of the system, is a subset of W T . In this paper we consider dynamical systems Σ = (Z + , R q , B), where R is the ring Z p r . Here p is a prime number and r is a positive integer. We study the theory of representations of these systems, in particular kernel representations, see also [7], [5]. For r ≥ 2 the ring Z p r is not a field. All multiples of p in Z p r are zero divisors and this induces several difficulties. Classical fundamental results for systems over a field are open problems for systems over the ring Z p r . One of these open problems is the development of a theory of row reduced representations and accompanying parametrization results. For behavioral systems over fields there exists a welldeveloped theory of representations, see e.g. [11], [14], [15], [16]. We define σ, the backward shift operator, acting on elements in W T as (σw)(k) = w(k + 1). Any behavior over a field that is linear, σ-invariant and complete (i.e., closed in the topology of point wise convergence) admits a kernel representation, that is, a representation of the form R(σ)w = 0, where R(ξ) is a polynomial matrix in the indeterminate ξ. As an example, for the system Σ = (Z + , R, B) with B = span {(3, 3, 3, · · · )} a kernel representation is given by (σ − 1)w = 0. For polynomial matrices over a field F, the concept of row reducedness is alternatively formulated in terms of the predictable-degree property (terminology from Forney's paper [2]), which is defined below. Recall that the row degree of a row polynomial vector is defined as the maximum of the degrees of its components. Definition 1.2: Let the matrix R(ξ) ∈ F g×q [ξ] with row degrees d 1 , . . . , d g . R(ξ) is said to have the predictabledegree property if for any nonzero polynomial vectorThus the row degree of a(ξ)R(ξ) can be predicted from the degrees in a(ξ) and the row degrees of R(ξ). For the field case it is proven in [2] and in [4, that the above property is equivalent to the property that the leading row coefficient matrix of R(ξ) has full row rank, i.e., that R(ξ) is row reduced. See also [13]. This provides an easy test to establish whether a kernel representation has the predictabledegree property or not. Fur...