arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2407.15002cs.RO
收藏

GET-Zero:面向零样本具身泛化的图具身 Transformer

GET-Zero: Graph Embodiment Transformer for Zero-shot Embodiment Generalization

Stanford University(斯坦福大学)

Austin Patel, Shuran Song

AI总结:

本文提出 GET-Zero,通过具身图结构注意力、行为克隆和自建模损失学习硬件感知控制策略,在四指机械手变化配置下实现零样本具身泛化并较基线提升 20%。

摘要:

本文介绍了 GET-Zero,这是一种模型架构与训练流程,用于学习具备具身感知能力的控制策略,该策略能够在无需重新训练的情况下立即适应新的硬件变化。为此,我们提出了 Graph Embodiment Transformer(GET,图具身 Transformer),这是一种 Transformer 模型,它将具身图连接结构作为注意力机制中的一种可学习结构偏置加以利用。我们采用行为克隆,将来自具身专属专家策略的演示数据蒸馏到一个具备具身感知能力的 GET 模型中,该模型以机器人的硬件配置为条件来作出控制决策。我们在灵巧的手中物体旋转任务上进行了案例研究,使用四指机器人手的不同配置,包括移除关节以及延长连杆长度。将 GET 模型与自建模损失结合使用,使 GET-Zero 能够零样本泛化到未见的图结构和连杆长度变化,相比基线方法取得了 20% 的提升。所有代码和定性视频结果均位于 https://get-zero-paper.github.io。

英文摘要:

This paper introduces GET-Zero, a model architecture and training procedure for learning an embodiment-aware control policy that can immediately adapt to new hardware changes without retraining. To do so, we present Graph Embodiment Transformer (GET), a transformer model that leverages the embodiment graph connectivity as a learned structural bias in the attention mechanism. We use behavior cloning to distill demonstration data from embodiment-specific expert policies into an embodiment-aware GET model that conditions on the hardware configuration of the robot to make control decisions. We conduct a case study on a dexterous in-hand object rotation task using different configurations of a four-fingered robot hand with joints removed and with link length extensions. Using the GET model along with a self-modeling loss enables GET-Zero to zero-shot generalize to unseen variation in graph structure and link length, yielding a 20% improvement over baseline methods. All code and qualitative video results are on https://get-zero-paper.github.io

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑