Terrain-Indexed Motion Priors and Geometric Support for Extreme Humanoid Parkour
Terrain-Indexed Motion Priors and Geometric Support for Extreme Humanoid Parkour
This paper presents a progressive imitation framework that learns terrain-appropriate humanoid parkour skills and adapts hand-knee support. Using demonstrations, the method employs privileged imitation for executable references, terrain-indexed adversarial motion priors for skill acquisition, and geometric support guidance for contact adaptation. On the Noetix E1, the policy achieves 96.34% hybrid success in simulation and transfers without real-world fine-tuning, completing 29 of 30 obstacle trials and four of five hybrid courses.
The study demonstrates a progressive imitation framework on the Noetix E1 humanoid that achieves 96.34% hybrid success in simulation and transfers to real-world trials without policy fine-tuning, completing 29 of 30 obstacle trials and four of five hybrid courses.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Environment
- SimulationOpen source S4
Reported numbers
trajectories
29
Reported trials
View original evidence
on, and velocity commands without an external skill label. On the Noetix E1, the policy achieves 96.34\% hybrid success in simulation and transfers without real-world policy fine-tuning, completing 29 of 30 individual obstacle trials and four of five hybrid courses. Support-adaptation ablations show higher traversal success for the complete adaptation stage on the tested terrains, while qualitative observations show
Open source S4trajectories
5
Reported trials
View original evidence
on, and velocity commands without an external skill label. On the Noetix E1, the policy achieves 96.34\% hybrid success in simulation and transfers without real-world policy fine-tuning, completing 29 of 30 individual obstacle trials and four of five hybrid courses. Support-adaptation ablations show higher traversal success for the complete adaptation stage on the tested terrains, while qualitative observations show
Open source S4trajectories
30
Reported trials
View original evidence
on, and velocity commands without an external skill label. On the Noetix E1, the policy achieves 96.34\% hybrid success in simulation and transfers without real-world policy fine-tuning, completing 29 of 30 individual obstacle trials and four of five hybrid courses. Support-adaptation ablations show higher traversal success for the complete adaptation stage on the tested terrains, while qualitative observations show
Open source S4trajectories
4
Reported trials
View original evidence
on, and velocity commands without an external skill label. On the Noetix E1, the policy achieves 96.34\% hybrid success in simulation and transfers without real-world policy fine-tuning, completing 29 of 30 individual obstacle trials and four of five hybrid courses. Support-adaptation ablations show higher traversal success for the complete adaptation stage on the tested terrains, while qualitative observations show
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
on, and velocity commands without an external skill label. On the Noetix E1, the policy achieves 96.34\% hybrid success in simulation and transfers without real-world policy fine-tuning, completing 29 of 30 individual obstacle trials and four of five hybrid courses. Support-adaptation ablations show higher traversal success for the complete adaptation stage on the tested terrains, while qualitative observations show
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org