November 19, 2021, zoom

While the potential of learning-based control paradigms is being demonstrated in numerous application domains, there are many open questions related to their safety, reliability, performance or also their computational properties limiting their wide-spread application. In this talk, I will discuss how predictive control concepts can be used to ensure safety in the form of satisfying a set of critical safety constraints when learning in a closed-loop dynamical system. Two directions will be discussed. On the one hand, predictive controllers with guarantees can directly benefit from learning to improve the description of any problem component, i.e. the model or objective, with the goal of enhancing closed-loop performance. On the other hand, predictive control principles can be used to enrich the safety properties of other learning-based control techniques via so-called predictive safety filters - a modular approach to guaranteeing safety via a scalable optimization-based formulation. Throughout the talk, the ideas will be highlighted using demonstrations of autonomous vehicles and robotics.

Melanie Zeilinger is an Assistant Professor at the Department of Mechanical and Process Engineering at ETH Zurich, Switzerland where she leads the Intelligent Control Systems group. She received the Diploma degree in engineering cybernetics from the University of Stuttgart, Germany, in 2006, and the Ph.D. degree with honors in electrical engineering from ETH Zurich, Switzerland, in 2011. From 2011 to 2012 she was a Postdoctoral Fellow with the Ecole Polytechnique Federale de Lausanne (EPFL), Switzerland. She was a Postdoctoral Researcher with the Max Planck Institute for Intelligent Systems, Tübingen, Germany until 2015 and with the Department of Electrical Engineering and Computer Sciences at the University of California at Berkeley, CA, USA, from 2012 to 2014. From 2018 to 2019 she was a professor at the University of Freiburg, Germany. Her awards include the ETH medal for her PhD thesis, a Marie-Curie IO fellowship and an SNF Professorship grant. She is one of the organizers of the new Conference on Learning for Dynamics and Control (L4DC). Her research interests include safe learning-based control with applications to robotics and human-in-the-loop control.