Metraipolis › archive › dc_8118e22692ef

Learning Foresight without Explicit Trajectories for 3D Diffusion Policies — arXiv:2609.20669

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.20669
License (as archived)
CC0-1.0
Archived
2026-10-10T05:12:22.782Z
SHA-256 of the archived text
1d505572f8c6793e731562702596ccd85212369b074b4d36286fe9510cc3ea06
Stable URL of this page
https://metraipolis.com/documento/dc_8118e22692ef

How to cite

Learning Foresight without Explicit Trajectories for 3D Diffusion Policies — arXiv:2609.20669 — arXiv (en), https://arxiv.org/abs/2609.20669 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8118e22692ef (Metraipolis archive dc_8118e22692ef, sha256 1d505572f8c6793e731562702596ccd85212369b074b4d36286fe9510cc3ea06)
BibTeX
@misc{dc8118e22692ef,
  title = {Learning Foresight without Explicit Trajectories for 3D Diffusion Policies — arXiv:2609.20669},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.20669},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8118e22692ef; sha256 1d505572f8c6793e731562702596ccd85212369b074b4d36286fe9510cc3ea06},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.20669. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

3D diffusion policies are strong at generating geometrically grounded actions from current observations, but successful manipulation requires not only knowing what motion is feasible now, but also anticipating where the interaction is heading. Existing policies largely leave such foresight to emerge implicitly from action learning. We introduce Movement Trend Guidance, a simple but effective way to provide this foresight without introducing an explicit plan. From a short observation history, the policy learns a compact latent representation of interaction evolution. During training, sparse future gripper states supervise this representation; at inference, only the latent is retained as future-oriented conditioning alongside the current observation. The latent provides global conditioning for action generation, while an additional gated FiLM branch is used only at the UNet bottleneck. Despite adding only 3.52% more parameters to DP3, our method preserves the original dense-action and receding-horizon formulation and consistently improves upon DP3 across RoboTwin2.0, LIBERO-40, and DexArt. It reaches 62.8% vs. 56.1% in 50-task RoboTwin2.0 mixed training, 71.93% vs. 37.08% on LIBERO-40, and 72.0% vs. 49.0% on five real-robot tasks. These results show that a diffusion policy can benefit substantially from knowing where an interaction is heading, without being told exactly where to move.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-18 in cs.CV).
Categorie arXiv: Robotics (cs.RO) ; Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.20669 — 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 2026-10-10T04:26:09.658Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).