Selective Residual Action Chunking for Vision-Guided Multi-Task Robotic Manipulation
Sensors · 05.11.2026 · 26 (21) · 6992
DOI: 10.3390/s26216992 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, online erschienen am 05.11.2026. Verfasst von Yibo Cao und 2 weiteren Autor:innen. Beteiligte Einrichtung: School of Artificial Intelligence, South China Normal University, Foshan 528225, China. Die Arbeit nennt 25 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Scientific Research Foundation of the Graduate School, South China Normal University, Key Technology Research (Second Batch) Program of Ningbo.
Abstract (Originalsprache)
Vision-guided robotic manipulation requires control actions that reflect visual observations, robot state, and task goals. Action Chunking with Transformers (ACT) performs well on individual tasks, but shared multi-task policies must accommodate different action distributions and state-dependent control requirements. We propose an ACT-based selective residual action-chunking policy using global and wrist RGB images together with proprioceptive feedback. A task-conditioned ACT model is first trained on multi-task demonstrations. Its parameters are then frozen while bounded residual experts learn candidate action corrections. A gain predictor uses observation-conditioned decoder features, robot state, task information, and base-action statistics to select a correction or retain the base prediction. Gain supervision is derived from reductions in imitation error. The robot executes an action prefix and updates its prediction from new observations, incorporating selective correction into closed-loop manipulation. Across all ten LIBERO-Goal tasks, the method achieves a mean success rate of 91.2%, exceeding MTACT and BAKU by 11.4 and 4.2 percentage points. On three jointly learned real-world tasks, it achieves 78.7%, exceeding MTACT by 10.0 percentage points. These results support selective residual action generation for vision-guided multi-task robot control.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- School of Artificial Intelligence, South China Normal University, Foshan 528225, China.
ORCID-Profile: Jiahong Huang
Förderung
- Scientific Research Foundation of the Graduate School, South China Normal University
- Key Technology Research (Second Batch) Program of Ningbo
Zitieren
APA
Cao, Y., Huang, J., & Yin, Z. (2026). Selective Residual Action Chunking for Vision-Guided Multi-Task Robotic Manipulation. Sensors, 26(21), 6992. https://doi.org/10.3390/s26216992
Vancouver
Cao Y, Huang J, Yin Z. Selective Residual Action Chunking for Vision-Guided Multi-Task Robotic Manipulation. Sensors. 2026;26(21):6992. doi:10.3390/s26216992
BibTeX
@article{Cao2026,
author = {Cao, Yibo and Huang, Jiahong and Yin, Zijun},
title = {Selective Residual Action Chunking for Vision-Guided Multi-Task Robotic Manipulation},
journal = {Sensors},
year = {2026},
volume = {26},
number = {21},
pages = {6992},
doi = {10.3390/s26216992}
}
EUVIMED