Computing the greatest X-eigenvector of a matrix in max-min algebra

Abstract

A vector $x$ is said to be an eigenvector of a square max-min matrix $A$ if $A\otimes x=x$. An eigenvector $x$ of $A$ is called the greatest $\textit{\textbf{X}}$-eigenvector of $A$ if $x\in\textit{\textbf{X}}=\{x; {\underline x}\leq x\leq {\overline x}\}$ and $y\leq x$ for each eigenvector $y\in\textit{\textbf{X}}$. A max-min matrix $A$ is called strongly $\textit{\textbf{X}}$-robust if the orbit $x,A\otimes x, A^2\otimes x,\dots$ reaches the greatest $\textit{\textbf{X}}$-eigenvector with any starting vector of $\textit{\textbf{X}}$. We suggest an $O(n^3)$ algorithm for computing the greatest $\textit{\textbf{X}}$-eigenvector of $A$ and study the strong $\textit{\textbf{X}}$-robustness. The necessary and sufficient conditions for strong $\textit{\textbf{X}}$-robustness are introduced and an efficient algorithm for verifying these conditions is described.