Metraipolis › archive › dc_171be8a36390

Sample, Simulate, Select: Physics-in-the-Loop Text-to-Motion for Humanoids Without Training — arXiv:2609.26420

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.26420
License (as archived)
CC0-1.0
Archived
2026-10-08T16:43:11.494Z
SHA-256 of the archived text
b880cde9ac8fcf6b8d24bee768592e6ff90c0b268d8eead44adfcf4a5e02c88b
Stable URL of this page
https://metraipolis.com/documento/dc_171be8a36390

How to cite

Sample, Simulate, Select: Physics-in-the-Loop Text-to-Motion for Humanoids Without Training — arXiv:2609.26420 — arXiv (en), https://arxiv.org/abs/2609.26420 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_171be8a36390 (Metraipolis archive dc_171be8a36390, sha256 b880cde9ac8fcf6b8d24bee768592e6ff90c0b268d8eead44adfcf4a5e02c88b)
BibTeX
@misc{dc171be8a36390,
  title = {Sample, Simulate, Select: Physics-in-the-Loop Text-to-Motion for Humanoids Without Training — arXiv:2609.26420},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.26420},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_171be8a36390; sha256 b880cde9ac8fcf6b8d24bee768592e6ff90c0b268d8eead44adfcf4a5e02c88b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.26420. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Text-to-motion models generate plausible human motion but do not model a robot's dynamics; whole-body tracking controllers execute robot references reliably but cannot replan an infeasible one. Recent language-to-humanoid systems bridge this gap by training. We measure how much of the gap closes with no training at all, by putting the deployment controller itself in the loop. Sample-simulate-select (S$^3$) draws $N$ motions per prompt from a frozen text-to-motion model, retargets each to a Unitree G1 by direction-matching inverse kinematics, rolls all of them out under full rigid-body dynamics with the pretrained SONIC tracking policy, and keeps the candidate the policy executed best. Because the verifier is the deterministic simulator itself, S$^3$ attains the any-of-$N$ ceiling by construction; what we measure is where that ceiling lies and what falls short of it. On 200 stratified HumanML3D test prompts with $N=8$, upright execution rises from 83.5% to 89.5% and hardware-gate passes from 33 to 85; on the complete test split (4,184 prompts) it rises from 80.5% to 89.5%. A kinematic verifier that predicts falls well (AUROC 0.90) recovers only a quarter of this gain: ranking a prompt's own candidates is harder than classifying the population. What selection cannot fix is one class, prompts that lower the pelvis, which a generator trained on retargeted robot data does execute. We further score the semantic fidelity of the executed motion with the standard text-motion evaluator, with a real-mocap control that attributes the loss to the robot projection, ablate the retargeter against GMR (complementary failures: the any-of-8 ceiling rises to 95.0% over both), and execute all 177 gate-selected clips on the real G1: every one completes standing, with hardware tracking error matching simulation ($r=0.94$).

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-23 in cs.CV).
Categorie arXiv: Robotics (cs.RO) ; Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.26420 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-08T15:56:18.247Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).