Metraipolis › archive › dc_f4e3037c665f

Mind the Phase: Effective Rank and Representation Health in Legged Locomotion — arXiv:2609.06958

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.06958
License (as archived)
CC0-1.0
Archived
2026-10-16T15:30:59.929Z
SHA-256 of the archived text
2741d0023deba83fb16913d74c05f226d11cced799350f96e3aa4324f3ec847c
Stable URL of this page
https://metraipolis.com/documento/dc_f4e3037c665f

How to cite

Mind the Phase: Effective Rank and Representation Health in Legged Locomotion — arXiv:2609.06958 — arXiv (en), https://arxiv.org/abs/2609.06958 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f4e3037c665f (Metraipolis archive dc_f4e3037c665f, sha256 2741d0023deba83fb16913d74c05f226d11cced799350f96e3aa4324f3ec847c)
BibTeX
@misc{dcf4e3037c665f,
  title = {Mind the Phase: Effective Rank and Representation Health in Legged Locomotion — arXiv:2609.06958},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.06958},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f4e3037c665f; sha256 2741d0023deba83fb16913d74c05f226d11cced799350f96e3aa4324f3ec847c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.06958. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement learning has become the leading paradigm in legged locomotion, enabling complex behaviors from backflips to parkour through massively parallel simulation. Under PPO's non-stationarity, shallow networks remain the de facto architecture, supported by carefully staged curricula and environments, yet the representations these policies learn stay poorly understood, leaving no training-time signal of how they will behave on hardware. In this work, we empirically study locomotion policies through the effective rank of the policy Jacobian and show that conditioning rank on the gait phase exposes architectural structure that global rank averages away. In particular, we find that standard architectural choices, namely layer normalization and residual connections, allocate roughly two more dimensions of effective rank to swing than to stance, which is fully absent in vanilla MLPs. Building on this, we propose a simple recipe that turns these representational signatures into smoother, more reliable sim-to-real transfer. In practice, this results in roughly 3x lower joint jitter that holds from simulation onto a physical Spot, suggesting that representation health is an effective training-time lens to track sim-to-real smoothness.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-09 in cs.RO).
Categorie arXiv: Robotics (cs.RO) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.06958 — 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 2026-10-16T15:11:47.559Z dalla pagina https://arxiv.org/catchup/cs.RO/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).