Reinforcement Learning and Optimal Control for Autonomous Systems I

(3 Credits)

This is the first part of a 2-semester course providing a full overview, assuming minimal prior background, of modern control methods from a learning and optimization perspective. In the first part of the course, we covers basics of optimization for sequential decision making problems (i.e. optimal control) as well as a half-semester introduction to reinforcement learning. The course includes topics such as Bellman’s principle of optimality, trajectory optimization, nonlinear model-predictive control, Q-learning and policy gradient methods. The course uses frequent competitions as projects to motivate the material, with students making an increasingly capable controller for a simulated robot over the course of the class. | Prerequisite: Knowledge of Python experience, multi-variable calculus