Gentle Manipulation Policy Learning via Demonstrations from VLM Planned Atomic Skills
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (22) · 18855-18863
DOI: 10.1609/aaai.v40i22.38955 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Jiayu Zhou und 5 weiteren Autor:innen.
Abstract (Originalsprache)
Autonomous execution of long-horizon, contact-rich manipulation tasks traditionally requires extensive real-world data and expert engineering, posing significant cost and scalability challenges. This paper proposes a novel framework integrating hierarchical semantic decomposition, reinforcement learning (RL), visual language models (VLMs), and knowledge distillation to overcome these limitations. Complex tasks are decomposed into atomic skills, with RL-trained policies for each primitive exclusively in simulation. Crucially, our RL formulation incorporates explicit force constraints to prevent object damage during delicate interactions. VLMs perform high-level task decomposition and skill planning, generating diverse expert demonstrations. These are distilled into a unified policy via Visual-Tactile Diffusion Policy for end-to-end execution. We conduct comprehensive ablation studies exploring different VLM-based task planners to identify optimal demonstration generation pipelines, and systematically compare imitation learning algorithms for skill distillation. Extensive simulation experiments and physical deployment validate that our approach achieves policy learning for long-horizon manipulation without costly human demonstrations, while the VLM-guided atomic skill framework enables scalable generalization to diverse tasks.
Zitieren
APA
Zhou, J., Wu, Q., Li, J., Chen, Z., Xiong, X., & Xu, R. (2026). Gentle Manipulation Policy Learning via Demonstrations from VLM Planned Atomic Skills. Proceedings of the AAAI Conference on Artificial Intelligence, 40(22), 18855-18863. https://doi.org/10.1609/aaai.v40i22.38955
Vancouver
Zhou J, Wu Q, Li J, Chen Z, Xiong X, Xu R. Gentle Manipulation Policy Learning via Demonstrations from VLM Planned Atomic Skills. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(22):18855-18863. doi:10.1609/aaai.v40i22.38955
BibTeX
@article{Zhou2026,
author = {Zhou, Jiayu and Wu, Qiwei and Li, Jian and Chen, Zhe and Xiong, Xiaogang and Xu, Renjing},
title = {Gentle Manipulation Policy Learning via Demonstrations from VLM Planned Atomic Skills},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {22},
pages = {18855--18863},
doi = {10.1609/aaai.v40i22.38955}
}
EUVIMED