Feedback realization of nonsingular precompensators for linear systems with delays (Q4346325)

The problem of implementing dynamic precompensators for linear systems with delays by realizable (i.e. nonanticipative) feedbacks is solved. A general model of linear time-invariant system with commensurate delays in states, inputs and outputs is considered. One assumes some hypotheses which allow to write its state and output equations in the form \[ \Sigma \begin{cases} x(t)= A(\nabla)x(t) +B(\nabla) u(t) \\ y(t)= C(\nabla)x(t) \end{cases} \tag{1} \] where \(x(t)\in \mathbb{R}^n\), \(u(t)\in \mathbb{R}^m\), \(y(t)\in \mathbb{R}^p\) and \(A(\nabla)\), \(B(\nabla)\), \(C (\nabla)\) are matrices of appropriate dimensions over the ring \[ {\mathcal R}_u (\nabla)= \bigl\{p (\nabla)/q (\nabla) \in\mathbb{R}(\nabla) \mid q(\nabla) \text{ has a nonzero constant term} \bigr\}. \] Therefore \(\Sigma\) is a linear system over a ring and particularly it is a delay system if \(\nabla\) is the delay operator \(\nabla x(t)= x(t-h)\), \(h>0\). The ring \({\mathcal R}_c\) of causal rational functions in the variables \(s\) and \(\nabla\) is also considered: \[ \begin{multlined} {\mathcal R}_c= \bigl\{h(s, \nabla)= p(s, \nabla )/q (s,\nabla) \in\mathbb{R} (s,\nabla) \mid h\text{ is }s \text{-proper, }\\ q(s, \nabla) =q_0(\nabla) +\cdots +q_k(\nabla) s^k\text{ and } q_k(0)\neq 0\bigr\}.\end{multlined} \] The problem of realization of a precompensator is the following: given a causal nonsingular precompensator \(T_c(s,\nabla) \in {\mathcal R}_c^{m \times m}\) determine a feedback \[ u(t)=-F(s,\nabla) x(t)+ G(\nabla) v(t) \tag{2} \] where \(F(s,\nabla) \in{\mathcal R}_c^{m \times n}\) and \(G(\nabla)\in {\mathcal R}_c^{m \times m}\), such that \[ T_c(s,\nabla) =\biggl(I_m+ F(s,\nabla) \bigl(sI_n-A(\nabla \bigr))^{-1} B(\nabla) \biggr)^{-1} g(\nabla). \tag{3} \] The control \(u(t)\) given by (2) is called Static State Feedback (SSF) if \(F\) does not depend on \(s\) and it is called Dynamic State Feedback (DSF) otherwise. If \(x(t)\) is replaced by \(y(t)\) in (2), then the corresponding notions of Static Output Feedback (SOF) and Dynamic Output Feedback (DOF) are defined. Necessary and sufficient conditions for the existence of SSF, SOF and DSF are given. The problem of Dynamic output realization leads to a model matching problem which is solved by the authors in another article.