Strong convergence in the stochastic averaging principle (Q1340544)

Let \(\{F(x,s, \omega), s \geq 0\}\) be a system of \(R^ d\)-valued processes indexed by \(x \in R^ d\). Consider the random ordinary differential equation \[ \dot X^ \varepsilon (t) = F(X^ \varepsilon (t), t/ \varepsilon), \qquad X^ \varepsilon (0) = x_ 0, \] and the averaged equation \[ \dot x(t) = \overline F \bigl( x(t) \bigr), \qquad x(0) = x_ 0, \] assuming that the limit \(\overline F(x) = \lim_{T \to \infty} {1 \over T} \int_ 0^ TEF (x,t) dt\) exists for all \(x \in R^ d\). Under rather weak conditions it is proved that \[ \lim_{\varepsilon \to 0}\quad \max_{0 \leq t \leq 1} | X^ \varepsilon (t) - x(t) | = 0 \] almost surely.