Stability of linear switching systems and Markov-Bernstein inequalities for exponents

Abstract

We analyse the problem of stability of a continuous time linear switching system (LSS) versus the stability of its Euler discretization. It is well-known that the existence of a positive {\tau} for which the corresponding discrete time system with step size {\tau} is stable implies the stability of LSS. Our main goal is to obtain a converse statement, that is, to estimate the discretization step size {\tau} > 0 up to a given accuracy {\epsilon} > 0. This leads to a method of deciding the stability of continuous time LSS with a guaranteed accuracy. As the first step, we solve this problem for matrices with real spectrum and conjecture that our method stays valid for the general case. Our approach is based on Markov-Bernstein type inequalities for systems of exponents. We obtain universal estimates for sharp constants in those inequalities. Our work provides the first estimate of the computational cost of the stability problem for continuous-time LSS (though restricted to the real-spectrum case).