Reinforcement learning transfer based on subgoal discovery and subtask similarity | IEEE Journals & Magazine | IEEE Xplore