Adaptive Control & Reinforcement Learning 

General Information

Term: Fall 2020

Lectures: TBD  Office Hours: ECOT 248, Times: TBD

Grading: The grade will be based on the following criteria:

Course Ad: [pdf]     Syllabus: [pdf]


Course Description

Course contents: Dynamic programming, Policy Evaluation, Policy Improvement, Policy Iteration, Maximum Principle for optimal control, the Hamilton-Jacobi-Bellman equation, the linear quadratic regulator, neural networks, system identification, learning dynamics, conditions of persistence of excitation, approximate dynamic programming, excitation-based online approximate optimal control, Lyapunov-based stability theory. Examples and applications in the areas of robotics, cyber-physical systems, autonomous vehicles and transportation systems. Students will develop: (a) A solid understanding of the principles behind adaptive control and neuro-adaptive reinforcement learning in continuous-time, discrete-time, continuous spaces and discrete spaces. (b) A solid understanding of the most common reinforcement learning algorithms, as well as their theoretical and practical limitations. Students will be endowed with a fundamental background that will allow them to pursue systems-centered PhD paths or explore future opportunities in the automation, optimization, and control systems marketplace.

The class has no official textbook. However, some useful references that we will use, include:

Additional Material

Additional textbooks: