Seminar Learning-based legged locomotion

What it is about

The seminar covers learning-based, bipedal locomotion of robots: fundamentals of decision making (including Markov decision processes), historical and current learning-based approaches, and open research questions. In the end you will understand the principles of modern (deep) learning methods for controlling legged robots and be able to assess the advantages and disadvantages of different approaches.

What you will be able to do

• Foundations of decision making for learning-enabled robots
• Understanding of MDP formalism in the context of locomotion
• Overview of historical and current learning-based methods
• Critical evaluation of research articles and identification of open problems
• In-depth knowledge in a selected state-of-the-art topic

What the module consists of

• Seminar / VorlesungenIntroduction to the fundamentals of bipedal locomotion and formalization of the decision problem
• Literaturarbeit / PräsentationenSelection of a current research article, development of a report and presentation
• DiskussionJoint discussion of the presented articles in the course

Teaching method

• VorträgeConveying the fundamental concepts and theories at the beginning of the seminar
• Selbststudium von ForschungsartikelnIn-depth engagement with a selected state-of-the-art article
• Präsentationen und DiskussionenPresentation of results in the course and critical exchange