The Parti-Game Algorithm for Variable Resolution Reinforcement Learning in Multidimensional State-Spaces