Constructive solution of the common invariant cone problem (Q7316802)

This paper studies sets of small dimensional matrices with an invariant cone and extends our knowledge of Perron-Frobenius Theory. For matrix families consisting of at most a handful of matrices this paper aims to find and decide whether common invariant cones exist, do not exist, and cannot be decided algorithmically in general. Four algorithms are described in detail: a 'direct' one, a primal-dual algorithm, one involving polyhedra cone computations and a `minimal cone' algorithm. Their performance is illustrated through numerical experiments, with MATLAB implementations. Families of matrices with invariant cones are studied in combinatorics, in convex optimization, dynamical systems, control theory and in other areas. Knowledge of a common invariant cone for a given set of matrices can simplify computations involving spectral quantities such as the joint and lower spectral radii and Lyapunov exponents.\N\NOne task of this paper is to find an invariant cone for a given set of matrices if such exists. A broad survey of known results is offered as well. This problem is algorithmically undecidable in general. It can be NP-hard for families consisting of four matrices. A broad introduction into this area and a detailed explanation of the known results rounds out the paper. A chapter deals with the question when a minimal cone is polyhedral and how to find such polyhedrals algorithmically. Test results are given in Tables and how to run the four algorithms in Matlab is explained for the authors' test-runs. How to arrange and test these algorithms, several example computation Matlab codes of the authors are given at the end.