Ralf Römer

Robot Learning PhD Student @TUM | ETH | EPFL | Bosch

Munich, Bavaria, Germany

About

I am a PhD student at the Learning Systems and Robotics Lab at TUM, advised by Prof. Angela Schoellig. My research aims to enable robots to safely execute complex tasks in uncertain, dynamic environments. For this, I am combining tools from machine learning, robotics, mathematics, and control theory. Specific research topics I am working on include uncertainty quantification, safety guarantees, and continual learning for diffusion policies and vision-language-action models. I am always open to research collaborations, so feel free to reach out if you are interested in working together! If you are a student interested in conducting a research project with me, please send me an email with your research interests, CV, and transcript.

Experience

  • Visiting Researcher at ETH Zürich
    Apr 2026 - Present · 3 mos

    Visiting PhD student at the Learning and Adaptive Systems Group led by Prof. Andreas Krause. Working on VLA post-training.

  • PHD Student at Technical University of Munich
    Feb 2024 - Present · 2 yrs 5 mos

    I am conducting research on generative robot policies, including vision-language-action models (VLAs). My research interests include the theoretical properties of diffusion- and flow-based policies, uncertainty quantification, safety and continual learning. Advisor: Prof. Angela Schoellig

  • Research Associate at Learning Systems and Robotics Lab
    Dec 2023 - Present · 2 yrs 7 mos

  • RIG Bootcamp on Foundational Behavior Models at Robotics Institute Germany
    Nov 2025 - Nov 2025 · 1 mo

    Selected as one of 6 tutors from across Germany to push research projects on VLAs and foundation models for robotics with 15 fellow PhD students. Resulted in two ongoing collaborations, which we are currently preparing for submission.

  • Master Thesis Student at Learning Systems and Robotics Lab
    Mar 2023 - Oct 2023 · 8 mos

    I investigated the roles of control frequency and data for the stability and closed-loop performance of uncertain robotic systems. The work has been published at the American Control Conference 2024. Advisors: Lukas Brunke, Dr. SiQi Zhou, Prof. Angela Schoellig