Reinforcement Learning in Complex Environments Through Multiple Adaptive Partitions