The acquisition and self-improvement of novel motor skills is among the most important problems in robotics. Reinforcement learning and imitation learning are two different but complimentary machine learning approaches commonly used for learning motor skills.
I will discuss various learning techniques we developed that can handle complex interactions with the environment. Complexity arises from non-linear dynamics in general and contacts in particular, taking multiple reference frames into account, dealing with high-dimensional input data, etc. I will present our methods on multi-modal imitation learning, the use of models for reinforcement learning in these scenarios, and discuss the particular challenges posed by deep reinforcement learning for these applications. I will illustrate these concepts with benchmark tasks and real robot experiments like pushing objects, unscrewing a lightbulb, or extracting teeth.
Biography: Jens Kober is an assistant professor at the TU Delft, Netherlands. He worked as a postdoctoral scholar jointly at the CoR-Lab, Bielefeld University, Germany and at the Honda Research Institute Europe, Germany. He graduated in 2012 with a PhD Degree in Engineering from TU Darmstadt. For his research he received the annually awarded Georges Giralt PhD Award for the best PhD thesis in robotics in Europe.
From 2007-2012 he was working with Jan Peters as a masters student and subsequently as a PhD student at the Robot Learning Lab, Max Planck Institute for Intelligent Systems (formerly part of the MPI for Biological Cybernetics). He has been a visiting research student at the Advanced Telecommunication Research (ATR) Center, Japan and an intern at Disney Research Pittsburgh, USA. His research interests include robotics, machine learning, and control.