The ability to learn will be a key requirement for future intelligent systems (robots, autonomous vehicles, etc.), which are envisioned to act autonomously in complex and changing environments. A core research area at the Autonomous Motion Department (AMD) is learning for control. We combine techniques from machine learning, control theory, and optimization to develop intelligent control algorithms for the next generation of intelligent systems. In particular, we focus on the special requirements that real-time control systems pose for learning algorithms, such as guarantees for stability, robustness, and efficient computation.
While rigorous theory and mathematical analysis form the basis of our research, we validate our methods in experiments on physical robots. We have a number of state-of-the-art robotic platforms at AMD to study various aspects of autonomous robots (see photo).
We are continuously looking for outstanding students who are eager to do their Master thesis on a challenging research project in a highly dynamic research environment. We have a variety of possible projects available, ranging from very theoretical to practical, and covering different aspects of learning control and robotics. Examples of possible topics include adaptive and learning control for complex robots, non-parametric learning of dynamic models, model-based reinforcement learning, learning-based model predictive control, and Bayesian optimization for control.
Dr. Sebastian Trimpe (email@example.com); Intelligent Control Systems Group; group website: http://trimpe.is.tuebingen.mpg.de/
CLS Student Project (MPG ETH CLS)
Information, Computing and Communication Sciences
Engineering and Technology