Gerhard Neumann

Gerhard Neumann joined the IAS group as Post-Doc in November 2011 with focus on probabilistic policy search, motor skill learning with movement primitives, probabilistic planning, reinforcement learning for robotics and hierarchical skill acquisition. His scientific goal is to put all these fields together in order to enable robots to learn a rich set of movement skills, which allows a robot to live autonomously in a real-world environment. Gerhard concentrates on the robot table tennis domain as challenging benchmark setup of rich motor skills.

Before coming to Darmstadt, Gerhard did his Ph.D. at the Graz University of Technology (TUG) under the supervision of Wolfgang Maass. Gerhard started his Ph.D. studies in August 2005. During his Ph.D. he was involved in several nation-funded and European-Union funded projects which concentrated on reinforcement learning for robotics, biologically inspired robotics, neural motor control and probabilistic inference for motor planning. During his Ph.D., he also collaborated with Jan Peters, Marc Toussaint and Auke Ijspeert. His Thesis, "On Motor Skill Learning and Movement Representations for Robotics" concentrated on value-based algorithms for motor skill learning, learning with different movement representations and policy search algorithms. Gerhard finished his PhD in April 2012.

Gerhard was born in Graz, Austria. Before doing his PhD, he finished his studies in telematics at the TUG in the year 2005. Gerhard also developed the Reinforcement Learning Toolbox, a C++ software library for RL algorithms, as his Master Thesis, which was frequently used by other scientists.

Gerhard can be found on Google Citations and DLBP.

Research interests

• Using inference for control, planning and learning
• (Hierarchical) Bayesian Models
• Reinforcement Learning for robotics
• Robot Applications for Machine Learning
• Movement representations for motor skill learning
• Stochastic Optimal Control
• Biologically Inspired Robotics and Motor Control

Key References

1. Neumann, G.; Peters, J. (2009). Fitted Q-iteration by Advantage Weighted Regression, Advances in Neural Information Processing Systems 22 (NIPS 2008), Cambridge, MA: MIT Press  download [PDF]
2. Neumann, G.; Maass, W; Peters, J. (2009). Learning Complex Motions by Sequencing Simpler Motion Templates, Proceedings of the International Conference on Machine Learning (ICML2009)  download [PDF]
3. Neumann G. (2011). Variational Inference for Policy Search in Changing Situations, Proceedings of the International Conference on Machine Learning (ICML 2011)   download [PDF]

For all publications please see his Publication Page

Workshops

Gerhard and Elmar Rückert organized a workshop for the EU-project Amarsi on the topic "Hands-on Probabilistic Inference for Motor Control". The workshop was mainly for Ph.D. students inside the project but also open to public.

Personal Information

Please see his CV.

Contact Information

Mail: Gerhard Neumann, TU Darmstadt, FB Informatik, FG IAS, Hochschulstr. 10, 64289 Darmstadt
Office: Room E325, Robert-Piloty Gebaeude S2|02
email geri (at) robot-learning (dot) de
work ++49-6151-16-64534