An algorithm for the multi-input complex eigenvalue assignment problem. (Q1406270)

A simple algorithm is presented to solve the linear multi-input problem of control theory. Given (i) \(\dot x= Ax+ Bu\), \(A\in \mathbb{R}^{n\times n}\), \(B\in \mathbb{R}^{n\times m}\), \(x\in \mathbb{R}^n\), \(u\in \mathbb{R}^m\), the problem is to find \(F\in \mathbb{R}^{m\times n}\) such that \(A+ BF\) of the system \(\dot x= (A+ BF)x\) has eigenvalues \[ \{\lambda_1,\lambda_2,\dots, \lambda_m, \lambda_{m+1},\dots, \lambda_{2m}, \lambda_{2m+1},\dots, \lambda_n\}, \] where the first \(2m\) numbers are complex with their complex conjugates and the others are real. The algorithm has two stages: (1) \(A\), \(B\) are transformed to a standard form \((H,R)= (P^T AP, P^T BU)\) where \(H\) is an unreduced block Hessenberg matrix and \(R\in\mathbb{R}^{n\times m}\) has zero's \(i> m\), \(J= 1,\dots, m\), \(R= {C\choose 0}\). (2) Matrices \(M\), \(L\), \(\Lambda\) are formed by real operations such that \(H+ M= L^{-1}\Lambda L\) where \(\Lambda\) has the given eigenvalues, and \(M\) has zero's \(i> m\), \(j= 1,\dots, n\), \(M={N\choose 0}\). Choosing \(K= C^{-1}N\) the feedback matrix is \(F= UKP^T\). The method generalizes a method for real eigenvalues by \textit{M. Arnold} and \textit{B. N. Datta} [IEEE Trans. Autom. Control 35, No. 10, 1149--1152 (1990; Zbl 0732.93031)]. Two numerical examples are provided.