Category : Reinforcement Learning and Control ; ORAL presentation Improved Switching among Temporally Abstract Actions