A Robust Data-Driven Approach for Online Learning and Manipulation of Unmodeled 3-D Heterogeneous Compliant Objects
A Robust Data-Driven Approach for Online Learning and Manipulation of Unmodeled 3-D Heterogeneous Compliant Objects
复制标题
DOI：
10.1109/lra.2018.2863376
复制
发表时间：
2018-10-01
影响因子：
5.2
通讯作者：
Armand, Mehran
中科院分区：
文献类型：
--
作者：
Alambeigi, Farshid;Wang, Zerui;Armand, Mehran
我们提出了一种通用的数据驱动方法来解决在存在未知干扰的情况下操纵具有异构物理性质的三维（3-D）柔性对象（CO）的问题。在本研究中，我们不假设对CO的变形行为和干扰类型（例如，内部或外部）有先验知识。我们也不会对CO的物理性质（如形状、质量和刚度）施加任何限制。提出的最优迭代算法结合所提供的视觉反馈数据，同时学习和估计CO的变形行为，以实现预期的控制目标。为了证明我们的算法的能力和鲁棒性，我们制作了两个新的异质柔性模型，并在达芬奇研究套件上进行了实验。实验结果证明了所提出方法的适应性、鲁棒性和准确性，因此，它适用于涉及CO操纵的各种医疗和工业应用。
We present a generic data-driven method to address the problem of manipulating a three-dimensional (3-D) compliant object (CO) with heterogeneous physical properties in the presence of unknown disturbances. In this study, we do not assume a prior knowledge about the deformation behavior of the CO and type of the disturbance (e.g., internal or external). We also do not impose any constraints on the CO's physical properties (e.g., shape, mass, and stiffness). The proposed optimal iterative algorithm incorporates the provided visual feedback data to simultaneously learn and estimate the deformation behavior of the CO in order to accomplish the desired control objective. To demonstrate the capabilities and robustness of our algorithm, we fabricated two novel heterogeneous compliant phantoms and performed experiments on the da Vinci Research Kit. Experimental results demonstrated the adaptivity, robustness, and accuracy of the proposed method and, therefore, its suitability for a variety of medical and industrial applications involving CO manipulation.