Filter-Aware Fine-Tuning for Safe Humanoid Whole-Body Tracking

What happened

arXiv:2610.02341v1 Announce Type: new Abstract: Safe whole-body motion is essential for deploying humanoid robots in unstructured environments. Runtime safety filters, such as control barrier functions (CBFs), offer a promising approach for enforcing newly introduced constraints via interventions on the tracker's outputs.

On Unitree G1 hardware, CoFiT reduces violation time by 83% for TWIST2 and completes every trial without operator intervention, whereas 50% of baseline trials require an operator stop. Modern humanoid control commonly separates reference specification from execution, with a planner, teleoperator, or motion generator providing a reference that a reinforcement-learning policy tracks through dynamically feasible whole-body control. Across diverse constraint scenes, CoFiT reduces violation time relative to filter-only training by 91% on TWIST2 and 21% on SONIC, while requiring smaller safety filter corrections.