Between MDPs and Semi-MDPs: A Framework for Temporal Abstraction in Reinforcement Learning

Options框架，分层强化学习时间抽象奠基。