Model-free optimal tracking policies for Markov jump systems by solving non-zero-sum games.