Alexander von Rohr

Institute for Data Science in Mechanical Engineering, RWTH Aachen University.


Building: 6050, Room: A 2.11

Dennewartstraße 27

52068 Aachen, Germany

I am a PhD student at RWTH Aachen University and my research is at the intersection between machine learning and robust control. The main topic of my dissertation is learning probabilistic robust controllers (a.k.a. policies), that are provably robust against uncertainty inherent in learning from data. I am also actively interested in Bayesian optimization in the context of reinforcement learning and controller tuning.

The first part of my PhD took place the Max Planck Institute for Intelligent Systems where I was a member of the Intelligent Control Systems Group lead by Prof. Sebastian Trimpe. In 2020 our research group moved to the RWTH Aachen University where we are in the process of building the new Institute for Data Science in Mechanical Engineering. I am an associated student of the International Max Planck Research School for Intelligent Systems IMPRS-IS and my PhD is supported by IAV.

Before joining the Max-Planck Institute for my master thesis and later my PhD in 2018 I studied Computer Science at the University of Lübeck. I obtained my Bachelor’s degree in Electrical Engineering at the BHT Berlin in 2013. In between the two degrees I worked as a full time Software Engineer in Hamburg.


Aug 24, 2022 Today we published Event-Triggered Time-Varying Bayesian Optimization on arXiv. The paper is about a finding the optimum of a time-varying objective function.
Jul 15, 2022 Our two submissions Improving the Performance of Robust Control through Event-Triggered Learning and On Controller Tuning with Time-Varying Bayesian Optimization for the IEEE Conference on Decision and Control (CDC) were accepted. We will present the papers at the Invited Session on Learning-​based Control.
Sep 28, 2021 Our paper on Local policy search with Bayesian optimization got accepted at NeurIPS 2021 😊.
Jun 23, 2021 Yesterday we published a preprint on Local policy search with Bayesian optimization. It is about active sampling for policy gradients without access to a first-order oracle.
Jun 7, 2021 I presented our paper Probabilistic robust linear quadratic regulators with Gaussian processes at the poster session of the the 3rd Annual Learning for Dynamics & Control Conference. Thanks to everyone who stopped by 😀.

selected publications

  1. Probabilistic robust linear quadratic regulators with Gaussian processes
    Alexander von Rohr, Matthias Neumann-Brosig, and Sebastian Trimpe
    In Proceedings of the 3rd Conference on Learning for Dynamics and Control 2021
  2. Local policy search with Bayesian optimization
    Sarah Müller*, Alexander von Rohr*, and Sebastian Trimpe
    In Advances in Neural Information Processing Systems 2021