On the Liapunov exponents of a linear system with Markov coefficients (Q789372)

Let \(\xi_ t\) be a homogeneous strong ergodic Markov process on a measurable space \({\mathfrak U}\). Consider the controlled stochastic dynamical system \[ \dot x_ t=(A+\epsilon B(\xi_ t))X_ t \] where A, B(u), \(u\in U\), are given matrices and \(\epsilon\) is a small parameter. Under some conditions on the spectrum of the matrix A the author shows that for all small values of \(\epsilon\) the expectation \(EX_ t\) can be represented in the form \[ EX_ t=\exp(Kt)(C+O(t))EX_ 0 \] where K, C are constant matrices, det \(C\neq 0\), and the residual term satisfies O(t)\(\to 0\) as \(t\to \infty\). A similar result is also obtained for the second moment \(EX_ tX_ t^{^*}\).