Overview

How can legged robots maintain stability on dynamically unstable surfaces? While quadrupeds excel on rough, rigid terrain, their performance on vertically oscillating ground is an underexplored problem. We trained quadruped gaits on an oscillating bridge in simulation, producing policies that transfer zero-shot to the real bridge with better stability and tracking than baselines.

Acknowledgments

This work was funded by German Research Foundation (DFG) under the HUMVIB project (grant number: 446124066) and the INTENTION project (grant number: 506123304). This project was also partially funded by National Science Centre, Poland under the OPUS call in the Weave program UMO-2021/43/I/ST6/02711.

Furthermore, we sincerely thank the German Research Center for AI (DFKI), Research Department: Systems AI for Robot Learning, for lending us the Unitree Go2 quadruped.

Citation

@article{stasica2025bridge,
    title={Bridge the Gap: Enhancing Quadruped Locomotion with Vertical Ground Perturbations},
    author={Stasica, Maximilian and Bick, Arne and Bohlinger, Nico and Mohseni, Omid and Fritzsche, Max Johannes Alois and H{\"u}bler, Clemens and Peters, Jan and Seyfarth, Andr{\'e}},
    journal={International Conference on Intelligent Robots and Systems},
    year={2025}
}