Jonas Frey

I am a PostDoc at Stanford (Autonomous System Lab) with Prof. Marco Pavone and UC Berkeley (BAIR) with Prof. Jitendra Malik. My research focuses on Learning Perception and Navigation for Legged Robots with a focus on training RL policies and Large Behaviour Models.

I obtained my Ph.D. in robotics at the Robotic Systems Lab at ETH Zurich and the Max Planck Institute for Intelligent Systems, co-advised by Prof. Marco Hutter and Prof. Georg Martius. During my Ph.D. I joined NASA’s Jet Propulsion Laboratory as a visiting researcher and continuously collaborate with the Dynamic Robotic Systems Group by Prof. Maurice Fallon at University of Oxford. I worked as a Collaborative Robotics Engineer at SEW-Eurodrive before starting my Master’s in Robotics Systems and Control at ETH Zurich. Throughout my Master’s I was supervised by Prof. Roland Siegwart and spent most of my time at the Autonomous Systems Lab. During my Bachelor’s, I worked at the chair for High Performance Humanoid Technologies with Prof. Tamim Asfour (KIT, Germany) and conducted my Bachelor’s Thesis with Prof. Takamitsu Matsubara in Japan (NAIST, Japan).

contact

Feel free to reach out to me via email or LinkedIn. If you are a student looking for project opportunities, I would greatly appreciate it if you could make the process smoother by including the following in your message:

Your project ideas or areas of interest
Your transcript of records
Your CV Looking forward to hearing from you!

news

Sep 01, 2025	I started my PostDoc at Stanford (Autonmous System Lab) with Prof. Marco Pavone and UC Berkeley (BAIR) with Prof. Jitendra Malik.
May 01, 2025	I’ve officially finished my Ph.D., supervised by Marco Hutter, Georg Martius, and Cesar Cadena—deep thanks to my committee members and long-term collaborators Maurice Fallon and Shehryar Khattak for their support over the years.
Apr 20, 2024	Our paper “Resilient Legged Local Navigation: Learning to Traverse with Compromised Perception End-to-End,” has been nominated for the Best Paper Award for ICRA2024! Key Insights: Unreliable Perception Systems: It’s impossible to always guarantee the reliability of perception systems. Need for Advanced Planning: There is a critical need for planners that can detect failures in perception systems and react appropriately. Use of Proprioceptive Data: Detecting these failures is feasible through the use of proprioceptive data. Limitations of Traditional Methods: Classical planning methods often fail to utilize such data effectively, prompting the need for innovative solutions. Using RL to Discover Emerging Navigation Strategies: Rather than relying on handcrafted heuristics, we use reinforcement learning (RL) and expose the robot to an environment with various types of perception failures. Our asymmetric actor-critic framework discovers emerging navigation strategies.
Mar 19, 2024	We’re thrilled to announce that our grant application for “Fostering Research on Mobile Robotics with High-Quality Data and Open Tooling” has been accepted! We aim to provide researchers with tailored, high-quality data for legged robots, to establish standardized benchmarks and foster the development of future algorithms. Stay tuned for updates as we advance mobile robotics research. Special thanks to Turcan Tuna, Marco Hutter, and Cesar Cadena! Learn more about the open research data initiative here. link
Dec 06, 2021	Our work: “Continual Learning of Semantic Segmentation using Complementary 2D-3D Data Representations” received the Best Paper Runner-Up Award at NeurIPS - 2021 4th Robot Learning Workshop Self-Supervised and Lifelong Learning link

selected publications

Locomotion policy guided traversability learning using volumetric representations of complex environments

Jonas Frey, David Hoeller, Shehryar Khattak, and 1 more author

In IROS, 2022

Bib HTML Video

@inproceedings{frey2022locomotion,
  author = {Frey, Jonas and Hoeller, David and Khattak, Shehryar and Hutter, Marco},
  booktitle = {IROS},
  iros = {true},
  organization = {IEEE},
  pages = {5722--5729},
  title = {Locomotion policy guided traversability learning using volumetric representations of complex environments},
  url = {https://arxiv.org/abs/2203.15854},
  year = {2022}
}

Fast Traversability Estimation for Wild Visual Navigation

Jonas Frey, Matias Mattamala, Nived Chebrolu, and 3 more authors

In Robotics: Science and Systems 2023, 2023

Bib HTML Video Code Poster Website

@inproceedings{frey2023fast,
  author = {Frey, Jonas and Mattamala, Matias and Chebrolu, Nived and Cadena, Cesar and Fallon, Maurice and Hutter, Marco},
  booktitle = {Robotics: Science and Systems 2023},
  rss = {true},
  title = {Fast Traversability Estimation for Wild Visual Navigation},
  url = {https://arxiv.org/abs/2305.08510},
  year = {2023}
}

RoadRunner - Learning Traversability Estimation for Autonomous Off-road Driving

Jonas Frey, Shehryar Khattak, Manthan Patel, and 5 more authors

IEEE Transactions on Field Robotics 2024, 2024

Bib HTML

@article{frey2024roadrunner,
  author = {Frey, Jonas and Khattak, Shehryar and Patel, Manthan and Atha, Deegan and Nubert, Julian and Padgett, Curtis and Hutter, Marco and Spieler, Patrick},
  journal = {IEEE Transactions on Field Robotics 2024},
  tfr = {true},
  title = {RoadRunner - Learning Traversability Estimation for Autonomous Off-road Driving},
  url = {https://arxiv.org/abs/2402.19341},
  year = {2024}
}

Boxi: Design Decisions in the Context of Algorithmic Performance for Robotics

Jonas Frey, Turcan Tuna, Lanke Frank Tarimo Fu, and 8 more authors

Robotics: Science and Systems 2025, 2025

Bib HTML Code

@article{frey2025boxi,
  author = {Frey, Jonas and Tuna, Turcan and Fu, Lanke Frank Tarimo and Weibel, Cedric and Patterson, Katharine and Krummenacher, Benjamin and M{\"u}ller, Matthias and Nubert, Julian and Fallon, Maurice and Cadena, Cesar and others},
  journal = {Robotics: Science and Systems 2025},
  rss = {true},
  title = {Boxi: Design Decisions in the Context of Algorithmic Performance for Robotics},
  year = {2025}
}