Pointwise stability and feedback control of linear systems with noncommensurate time delays (Q2266693): Difference between revisions

A control system described by the following equations \[ \sum M_ i\dot x(t-h_ i)=\sum A_ ix(t-h_ i)+B_ iu(t-h_ i) \] \[ \sum N_ iy(t-h_ i)=\sum C_ ix(t-h_ i)+Du(t-h_ i) \] is considered where the time-dependent vectors x, u, y are real finite-dimensional, the coefficient matrices are real, the delays \(h_ i\) are nonnegative and the equations above are solvable with respect to \(\dot x(\)t) and y(t). After extracting noncommensurate generators for the delays the coefficient matrices are considered as over \({\mathbb{R}}[z]={\mathbb{R}}[z_ 1,...,z_ q]\) (we write M(z), A(z) etc.). Pointwise stability is understood as the property that det M(z)\(\neq 0\), det N(z)\(\neq 0\), and \(\det (sI-M^{-1}(z)A(z))\neq 0\) for s in the closed right halfplane and z in the closed unit polydisc. It implies, in particular, asymptotic stability for all nonnegative values of the generators of the delays. Among the main results of the paper is Thm. 2.26 characterizing pointwise stability in terms of asymptotic stability independent of delays and some additional algebraic conditions; Thm. 3.5 stating that pointwise stabilizability and pointwise detectability is a necessary and sufficient condition for the existence of a linear dynamic feedback pointwisely stabilizing the system; and Thm. 4.3 on equivalence between solvability of certain matrix Bezout equations and the existence of various types of realizations for strictly proper rational matrices.