Reinforcement Learning and Approximate Dynamic Programming for Feedback Control