Modern Adaptive Control and Reinforcement Learning

Drew Bagnell, Byron Boots, Sanjiban Choudhury

A set of practical tools to build decision making procedures for machines interacting with the world. Our applications vary from video games and web-search to robot manipulation and self-driving vehicles. Our focus is on intuition and informal mathematical argument to build that intuition, and on techniques we’ve seen work multiple times on hard decision making problems. We try to outline the techniques and ways of thinking we’d be most likely to pull out in practice. Throughout, we attempt to point to rigorous derivations and the original literature on the topics.

Current draft of the book: PDF