DiffWBC: Unlocking Steerable General Humanoid Motion Tracking via One-Step Diffusion Whole-Body Control
DiffWBC: Unlocking Steerable General Humanoid Motion Tracking via One-Step Diffusion Whole-Body Control
This paper introduces DiffWBC, a generative whole-body control framework for humanoid robots that unifies general motion tracking, test-time steering, and simulation-free failure discovery. The method utilizes a one-step diffusion model to achieve low local tracking error (20.1 mm on BONES) and zero-shot adaptation to various tasks including hurdle navigation and actuator faults without retraining or physics simulations.
The authors report DiffWBC, a diffusion-based whole-body control method that achieves the lowest local tracking error of 20.1 mm on the BONES benchmark among six general trackers and demonstrates zero-shot steering capabilities on a physical Unitree G1.
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.
Reported numbers
tracking error
20.1
View original evidence
mulation-free failure discovery within a single controller: (1) Generative general tracking: DiffWBC requires a single forward pass per 50 Hz control step and can be distilled from an RL teacher or trained from scratch via model-based RL, achieving the lowest local tracking error among six general trackers on held-out benchmarks (20.1 mm on BONES) and deploying on a physical Unitree G1. (2) Zero-shot test-time steeri
Open source S3Reported precision
91%
View original evidence
s (91% precision in top 2%) without physics simulations; targeted fine-tuning on flagged clips resolves 50% of failure cases. Videos and interactive demonstrations are available at https://diffwbc.github.io/#selfassess.
Open source S5
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
mulation-free failure discovery within a single controller: (1) Generative general tracking: DiffWBC requires a single forward pass per 50 Hz control step and can be distilled from an RL teacher or trained from scratch via model-based RL, achieving the lowest local tracking error among six general trackers on held-out benchmarks (20.1 mm on BONES) and deploying on a physical Unitree G1. (2) Zero-shot test-time steeri
Open source S3
ng: differentiable cost gradients and body-group masks modulate the frozen controller online, substantially outperforming kinematic reference editing (0.84 clean pass at 1.25 m ceiling vs. 0.10) and zero-shot generalizing across hurdles, spatial reaching, actuator faults, and tray balancing. (3) Simulation-free failure discovery: plan dispersion across generation noise effectively identifies physical tracking failure
Open source S4
s (91% precision in top 2%) without physics simulations; targeted fine-tuning on flagged clips resolves 50% of failure cases. Videos and interactive demonstrations are available at https://diffwbc.github.io/#selfassess.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org