On the selection of poles in the single-input pole placement problem (Q1970446)

An interesting issue in the theory of linear systems design is the sensitivity of the pole assignment problem. Consider the controllable pair \((A,b)\in \mathbb{C}^{n\times n}\times \mathbb{C}^n\), corresponding to the time-invariant system \(x'(t)= Ax(t)+ bu(t)\). Let \(\lambda= \{\lambda_1,\dots, \lambda_n\}\) be a collection of \(n\) complex numbers. The pole assignment problem is to find the gain matrix \(-f^\top\in \mathbb{C}^{1\times n}\) such that the spectrum of the closed-loop system matrix \(A- bf^\top\) is equal to \(\lambda\). Let \(\delta A\), \(\delta b\) be perturbations in \(A\), \(b\) such that the pair \([A+\delta A,b+\delta b)\) remains controllable. If \(f+\delta f\) is the solution of the perturbed problem, then it is necessary to estimate the norm \(\|\delta f\|\) of \(f\) as a function of \(\|\delta A\|\), \(\|\delta b\|\). The authors assume that the elements of \(\lambda\) are pairwise distinct. They then derive a perturbation estimate for \(\|\delta f\|\) (Theorem 1.1) and use an explicit expression (2.6) for \(f\). Finally, they propose a pole assignment algorithm. Reviewer's remarks: The results of Theorem 1.1, relation (2,6) and others heavily depend on the fact that the desired poles \(\lambda_i\) are distinct (the corresponding expressions do not exist when there are multiple poles, or they are very large when there are nearly multiple poles). At the same time, the correct estimates should not depend on the desired poles at all, see [\textit{M. Konstantinov}, \textit{P. Petkov} and \textit{N. Christov}, IEEE Trans. Autom. Control 42, 568-573 (1997; Zbl 0878.93020)]. In this paper a complete asymptotic as well as non-local perturbation analysis of the general problem of synthesis of output or state feedback is presented without any restrictions on the desired poles. This result seems not to be very popular in the mathematical literature since several weaker results have been published after it appeared in 1997.