Approximating Euclidean by Imprecise Markov Decision Processes