I have graduated and moved to Boston Dynamics in Cambridge, MA, USA. You can find more up-to-date information on my personal homepage.
Michael Lutter
Quick Info
Research Interests
Machine Learning, Robotics,
Reinforcement Learning,
Deep Learning
More Information
CV
Publications
Google Scholar
Contact Information
Mail. Michael Lutter
emailmichael@robot-learning.de

Michael Lutter joined the Institute for Intelligent Autonomous Systems (IAS) at TU Darmstadt in July 2017 and completed his Ph.D. in November 2021. Prior to this Michael held a researcher position at the Technical University of Munich (TUM) for bio-inspired learning for robotics. During this time he worked for the Neurorobotics subproject of the
Human Brain Project, a European H2020 FET flagship project. In addition, he taught the classes "Deep Learning for Autonomous Systems” and “Fundamentals of Computer Science for Neuroengineering” within the
Elite Master Programm Neuroengineering and participated in teaching
"Think. Make. Start.", a two-week prototyping course. His educational background covers a Bachelors in Engineering Management from University of Duisburg Essen and a Masters in Electrical Engineering from the Technical University of Munich. During his undergraduate studies he also spent one semester at the Massachusetts Institute of Technology studying electrical engineering and computer science. In addition to his studies, Michael worked for ThyssenKrupp, Siemens and General Electric and received multiple scholarships for academic excellence and his current research.
Research Topic
During his Ph.D. Michael works on a joint project with ABB Corporate Research on bringing robot learning to industrial applications. The current assembly automation using robotics is limited by the high cost and long duration of robot programming and the inflexible programs. Hence, robotic automation is only applicable to standardised and high volume manufacturing. However, the trend for highly personalised products requires assembly automation for smaller volumes. With other words, future robots must perform thousand tasks several times rather than a single task thousand times.
Robot learning promises the generation of intelligent robots capable of being programmed by workers rather than engineers and capable of transferring knowledge between tasks. Therefore, reducing cost for programming and time-to-deployment, which is a key enabler for robotic automation for low-volume and personalised manufacturing. Within this context, Michael first evaluates the current state of the art of robot learning by demonstrating the current capabilities on a small assembly task. Afterwards, he will focus his research on transferring knowledge between similar domain tasks.
Research Interests
Machine Learning:
Deep Learning, Non-Convex Optimisation, Reinforcement Learning, Structured Learning, Inductive Biases, Safe Exploration, Neuro-Inspired Learning, Neuromorphic Hardware
Robotics:
High-Speed Robotics, Industrial Manipulators, Humanoids, Dexterous Manipulation, Learning for Control, Optimal Control, Robust Control, Motion Representation
Key References
- Lutter, M.; Ritter, C.; Peters, J. (2019). Deep Lagrangian Networks: Using Physics as Model Prior for Deep Learning, International Conference on Learning Representations (ICLR).
Download Article [PDF] BibTeX Reference [BibTex]
- Lutter, M.; Mannor, S.; Peters, J.; Fox, D.; Garg, A. (2021). Robust Value Iteration for Continuous Control Tasks, Robotics: Science and Systems (RSS).
Download Article [PDF] BibTeX Reference [BibTex]
- Lutter, M.; Mannor, S.; Peters, J.; Fox, D.; Garg, A. (2021). Value Iteration in Continuous Actions, States and Time, International Conference on Machine Learning (ICML).
Download Article [PDF] BibTeX Reference [BibTex]
- Lutter, M.; Peters, J. (2019). Deep Lagrangian Networks for end-to-end learning of energy-based control for under-actuated systems, International Conference on Intelligent Robots and Systems (IROS).
Download Article [PDF] BibTeX Reference [BibTex]