On the Computation of Invariant Measures in Random Dynamical Systems

Invariant measures of dynamical systems generated e. g. by dierence equations can be computed by discretizing the originally continuum state space, and replacing the action of the generator by the transition mechanism of a Markov chain. In fact they are approximated by stationary vectors of these Markov chains. Here we extend this well known approximation result and the underlying algorithm to the setting of random dynamical systems, i.e. dynamical systems on the skew product of a probability space carrying the underlying stationary stochasticity and the state space, a particular non-autonomous framework. The systems are generated by dierence equations driven by stationary random processes modelled on a metric dynamical system. The approximation algorithm involves spatial discretizations and the denition of appropriate random Markov chains with stationary vectors converging to the random invariant measure of the system.