Learning modular language-conditioned robot policies through attention
Learning modular language-conditioned robot policies through attention
复制标题
DOI：
10.1007/s10514-023-10129-1
复制
发表时间：
2023-08
影响因子：
3.5
通讯作者：
Yifan Zhou;Shubham D. Sonawani;Mariano Phielipp;Heni Ben Amor-Heni-Ben Amor-2236830725;Simon Stepputtis
中科院分区：
文献类型：
--
作者：
Yifan Zhou;Shubham D. Sonawani;Mariano Phielipp;Heni Ben Amor-Heni-Ben Amor-2236830725;Simon Stepputtis
培训受语言制约的策略通常是耗时且资源密集的。此外，所得到的控制器是为特定的机器人量身定制的，因此很难将它们转移到具有不同动力学的其他机器人上。为了应对这些挑战，我们提出了一种称为分层模块化的新方法，该方法可以更有效地训练和随后在不同类型的机器人之间转移此类策略。该方法结合了监督注意，通过重用功能构建块，弥合了模块化和端到端学习之间的差距。在本文中，我们以之前的工作为基础，通过扩展层次结构以包含新任务和引入用于合成大量新对象的自动化管道，展示了扩展的实用程序和改进的性能。我们通过广泛的模拟和现实世界的机器人操作实验证明了这种方法的有效性。
Training language-conditioned policies is typically time-consuming and resource-intensive. Additionally, the resulting controllers are tailored to the specific robot they were trained on, making it difficult to transfer them to other robots with different dynamics. To address these challenges, we propose a new approach called Hierarchical Modularity, which enables more efficient training and subsequent transfer of such policies across different types of robots. The approach incorporates Supervised Attention which bridges the gap between modular and end-to-end learning by enabling the re-use of functional building blocks. In this contribution, we build upon our previous work, showcasing the extended utilities and improved performance by expanding the hierarchy to include new tasks and introducing an automated pipeline for synthesizing a large quantity of novel objects. We demonstrate the effectiveness of this approach through extensive simulated and real-world robot manipulation experiments.