See TUMonline

After completing the module, the students have an understanding of the basics of optimal control and reinforcement learning and an overview of the state-of-the-art of research for robotic applications. Also, students are able to identify the main research directions in the field and summarize the state-of-the-art in a specific research direction by investigating and analyzing research papers. In addition, they are capable of conducting group discussions and collaborative writing, and present fundamental principles, critical analysis and findings clearly.

Optimal Control and Reinforcement Learning for Robotics is a seminar course which will introduce the basics and explore advanced topics in optimal control and reinforcement learning. Students will first be introduced to the foundations of optimal control theory and reinforcement learning. Students will then be guided to explore advanced topics which focus on robotic applications.

- Fundamentals of control theory - Fundamentals of robotics

Students will be given introductory lectures and guided to conduct reading and discussions of research papers.

In this course, students are required to demonstrate their ability to do literature review of research papers and discuss the principles, state-of-the-art methods and their limitations in the field of Optimal Control (OC) and Reinforcement Learning (RL) for Robotics. The module grade is based on the completeness and quality of a written report (70%) and presentation (30%). The report is written in teamwork. It focuses on discussing several advanced topics in OC and RL with the focus being on the clear and precise summary of the most important facts and the conclusions drawn from those topics (15-20 page report). In the oral presentation followed by a discussion, the students show that they understand important applications, perspectives, and opportunities for selected technologies in the field of OC and RL. The presentation focuses on the visualization of the results and conclusive lines of argument. (20 minutes presentation, 20 minutes discussion afterward).

- A. Barto and R. Sutton, Reinforcement Learning: An Introduction - D. Bertsekas (2019), Reinforcement Learning and Optimal Control, Athena Scientific - D. Bertsekas (2017), Dynamic Programming and Optimal Control, 4th Edition - E. Todorov (2006), Optimal Control Theory (book chapter) - R. Tedrake (2021), Underactuated Robotics: Algorithms for Walking, Running, Swimming, Flying, and Manipulation (online lecture notes) - The most famous conferences and journals in the area of robotics, AI and machine learning such as ICRA, IROS, NIPS, The International Journal of Robotics Research, IEEE Transactions on Robotics, etc.