Sutton and Barto's book on reinforcement learning is available online at http://www.cs.ualberta.ca/~sutton/book/the-book.html, and is probably the definitive work on the subject.