EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Action Energy World Models Plan Long Horizon Robotic Cloth Folding Through Energy Landscapes and Planning

Ethan Lawson, Olivia Ellis, Samuel Hamilton

International Journal of Computational and Biological Sciences · 15.06.2026 · 3 (2) · 112-123

DOI: 10.66238/ijcbs90 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Computational and Biological Sciences, online erschienen am 15.06.2026. Verfasst von Ethan Lawson und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The autonomous manipulation of deformable objects, particularly the folding of garments and fabrics over extended temporal horizons, remains a formidable challenge in modern robotic research. Unlike rigid objects, which possess easily definable coordinate frames and predictable dynamic behaviors, cloth exhibits infinite degrees of freedom, severe self-occlusion during manipulation, and highly non-linear dynamics. This paper introduces a novel framework centered on Action-Energy World Models, which replaces traditional forward-predictive video or state models with an energy-based approach to sequence planning. By evaluating the scalar energy of state-action-state transitions in a highly compressed latent space, the proposed model identifies dynamically viable manipulation trajectories without requiring pixel-perfect reconstruction of the future cloth state. We present a comprehensive methodology detailing the integration of self-supervised latent encoding, energy-based transition modeling, and model predictive control optimized for long-horizon tasks requiring sequential geometric transformations. Through rigorous evaluation in simulated environments and real-world robotic setups, we demonstrate that the Action-Energy World Model significantly outperforms baseline predictive frameworks in multi-step folding tasks. The energy-based approach effectively mitigates compounding errors typically associated with long-horizon autoregressive predictions, ensuring robustness against perceptual noise and unexpected physical deformations. The findings suggest a paradigm shift in how robotic agents can conceptualize and plan interactions with complex, unstructured materials in domestic and industrial automation contexts.

Zitieren

APA

Lawson, E., Ellis, O., & Hamilton, S. (2026). Action Energy World Models Plan Long Horizon Robotic Cloth Folding Through Energy Landscapes and Planning. International Journal of Computational and Biological Sciences, 3(2), 112-123. https://doi.org/10.66238/ijcbs90

Vancouver

Lawson E, Ellis O, Hamilton S. Action Energy World Models Plan Long Horizon Robotic Cloth Folding Through Energy Landscapes and Planning. International Journal of Computational and Biological Sciences. 2026;3(2):112-123. doi:10.66238/ijcbs90

BibTeX
@article{Lawson2026,
  author = {Lawson, Ethan and Ellis, Olivia and Hamilton, Samuel},
  title = {Action Energy World Models Plan Long Horizon Robotic Cloth Folding Through Energy Landscapes and Planning},
  journal = {International Journal of Computational and Biological Sciences},
  year = {2026},
  volume = {3},
  number = {2},
  pages = {112--123},
  doi = {10.66238/ijcbs90}
}
Kostenlos starten →