Adaptive Action Chunking for Robotic Imitation Learning
Biomimetics · 02.05.2026 · 11 (5) · 316
DOI: 10.3390/biomimetics11050316 ↗Kurz gefasst
Fachartikel in der Zeitschrift Biomimetics, online erschienen am 02.05.2026. Verfasst von Qingpeng Wen und 5 weiteren Autor:innen. Beteiligte Einrichtung: School of Sino-German Robotics, Shenzhen University of Information Technology, Shenzhen 518172, China. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Projects of Shenzhen Institute of Information Technology, Open Fund of the Guangdong Key Laboratory of Electromagnetic Control and Intelligent Robot, Shenzhen University, Engineering Technology Research Center of Guangdong Province’s General Universities and Colleges.
Abstract (Originalsprache)
Action chunking strategies in robot imitation learning struggle to dynamically balance between long-range motion efficiency and short-range operational precision due to their fixed planning horizon. This paper presents an Adaptive Action Chunking framework that enables robots to dynamically predict the optimal action chunk length based on real-time visual context. We design an end-to-end dual-branch network comprising a shared visual encoder, a parallel action prediction head, and a chunk-size prediction head. Experiments on two real-world bimanual robot manipulation tasks (transport-and-place and flip-and-handover) demonstrate that the method autonomously derives two distinct intelligent strategy patterns—phase-aware switching and sustained high-frequency adjustment—in response to task uncertainty. It significantly outperforms fixed-chunk baselines in both success rate and efficiency. Ablation studies confirm that the performance gain stems from the adaptive decision-making mechanism itself.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- School of Sino-German Robotics, Shenzhen University of Information Technology, Shenzhen 518172, China.
- Shanghai AI NEV Innovative Platform Co., Ltd., Shanghai 201805, China.
- State Grid Shanghai Municipal Electric Power Company, Shanghai 200125, China.
ORCID-Profile: Yuepeng Zhang
Förderung
- Projects of Shenzhen Institute of Information Technology
- Open Fund of the Guangdong Key Laboratory of Electromagnetic Control and Intelligent Robot, Shenzhen University
- Engineering Technology Research Center of Guangdong Province’s General Universities and Colleges
Zitieren
APA
Wen, Q., Zhu, H., Zhang, Y., Xia, L., Gao, B., & Li, Z. (2026). Adaptive Action Chunking for Robotic Imitation Learning. Biomimetics, 11(5), 316. https://doi.org/10.3390/biomimetics11050316
Vancouver
Wen Q, Zhu H, Zhang Y, Xia L, Gao B, Li Z. Adaptive Action Chunking for Robotic Imitation Learning. Biomimetics. 2026;11(5):316. doi:10.3390/biomimetics11050316
BibTeX
@article{Wen2026,
author = {Wen, Qingpeng and Zhu, Haomin and Zhang, Yuepeng and Xia, Linzhong and Gao, Bo and Li, Zhuozhen},
title = {Adaptive Action Chunking for Robotic Imitation Learning},
journal = {Biomimetics},
year = {2026},
volume = {11},
number = {5},
pages = {316},
doi = {10.3390/biomimetics11050316}
}
EUVIMED