What does it take to teach a robot a new skill? Do we need massive datasets, sophisticated architectures, or just better ways to collect data? This seminar explores the rapid evolution of imitation learning—from early behavior cloning methods to the latest foundation models capable of generalizing across embodiments, tasks, and environments.
Over the semester, we will dive into 15 milestone papers that define the state of the art, covering topics such as:
- Transformer-based sequence modeling for policy learning
- Scalable, low-cost data collection strategies
- Diffusion-based policies and generative action models
- Vision-language-action (VLA) models as robotic foundation models
- Interactive learning through human corrections and hybrid RL approaches
This seminar is designed for master’s students with a background in robotics, machine learning, or artificial intelligence. Familiarity with the following topics will be helpful:
- Deep learning fundamentals (e.g., sequence models, transformers, diffusion models)
- Experience with robotics or robot learning frameworks is beneficial but not required
- Reinforcement learning basics
Rhythmus | Tag | Uhrzeit | Format / Ort | Zeitraum | |
---|---|---|---|---|---|
wöchentlich | Mo | 14-16 | CITEC 2.015 | 07.04.-18.07.2025 |
Modul | Veranstaltung | Leistungen | |
---|---|---|---|
39-M-Inf-AI-adv Advanced Artificial Intelligence | Advanced Artificial Intelligence: Seminar 1 | Studienleistung
|
Studieninformation |
39-M-Inf-ASE-adv Advanced Autonomous Systems Engineering | Advanced Autonomous Systems Engineering: Seminar 1 | Studienleistung
|
Studieninformation |
Die verbindlichen Modulbeschreibungen enthalten weitere Informationen, auch zu den "Leistungen" und ihren Anforderungen. Sind mehrere "Leistungsformen" möglich, entscheiden die jeweiligen Lehrenden darüber.
Each student will choose one paper to present in a 45-minute talk, followed by a group discussion which they lead.