Abstraction and Generalization in Reinforcement Learning: A Summary and Framework