An Atomic Skill Library Construction Method for Data-Efficient Embodied Manipulation

TL;DR

<3-5 sentence high-level summary> The paper tackles the data hunger and limited generalization of end-to-end embodied manipulation by proposing a three-wheeled, data-driven approach to construct a growing atomic skill library. It combines Vision-Language-Planning for task decomposition, a semantic abstraction step to define general atomic skills, and Vision-Language-Action fine-tuning to instantiate those skills, enabling dynamic library expansion with minimal data. Empirical results on real-world robots show that this method achieves comparable or better task success with less data and demonstrates strong transfer to new tasks and environments across different VLA backbones. The work offers a practical pathway to scalable, data-efficient embodied manipulation with rapid adaptation to unseen tasks.

Abstract

Embodied manipulation is a fundamental ability in the realm of embodied artificial intelligence. Although current embodied manipulation models show certain generalizations in specific settings, they struggle in new environments and tasks due to the complexity and diversity of real-world scenarios. The traditional end-to-end data collection and training manner leads to significant data demands. Decomposing end-to-end tasks into atomic skills helps reduce data requirements and improves the task success rate. However, existing methods are limited by predefined skill sets that cannot be dynamically updated. To address the issue, we introduce a three-wheeled data-driven method to build an atomic skill library. We divide tasks into subtasks using the Vision-Language-Planning (VLP). Then, atomic skill definitions are formed by abstracting the subtasks. Finally, an atomic skill library is constructed via data collection and Vision-Language-Action (VLA) fine-tuning. As the atomic skill library expands dynamically with the three-wheel update strategy, the range of tasks it can cover grows naturally. In this way, our method shifts focus from end-to-end tasks to atomic skills, significantly reducing data costs while maintaining high performance and enabling efficient adaptation to new tasks. Extensive experiments in real-world settings demonstrate the effectiveness and efficiency of our approach.

Figures (3)

• Figure 1: Three-Wheeled Self-Driven Atomic Skill Library Construction and Inference Pipeline.
• Figure 2: The VLP agent reasoning chain framework based on spatial intelligence information.
• Figure 3: Task definitions and visualization.