 |
| Georg Martius |
| organizer • speaker |
Bio
Georg Martius is leading a research group on Autonomous Learning at the Max Planck Institute for Intelligent Systems in Tübingen, Germany. Before joining the MPI in Tübingen, he was a postdoc fellow at the IST Austria in the groups of Christoph Lampert and Gašper Tkačik after being a postdoc at the Max Planck Institute for Mathematics in the Sciences in Leipzig. He pursues research in autonomous learning, that is how an embodied agent can determine what to learn, how to learn, and how to judge the learning success.
His research focus is on machine learning for robotics, including internal model learning, reinforcement learning, representation learning, differentiable reasoning and haptics.
Google scholar page • Personal webpage
Talk
I will present our recent work on combining intrinsic motivation signals with model-based planning methods to make robots learn by freely playing how to interact effectively with the world.
The learned world models enable zero-shot planning for new tasks, which is really exciting. Next, we consider how to impose more structure into the play phase. It turns out, we can operationalize the concept of regularity as an additional intrinsic motivation that leads to improved performance in downstream construction tasks.
The talk will also present the basic building blocks that we developed for our most recent results, namely fast online planning methods and uncertainty modeling.