Dexterous Cable Manipulation: Taxonomy, Multi-Fingered Hand Design, and Long-Horizon Manipulation

TL;DR

The paper addresses the gap between human dexterity and robotic cable manipulation by introducing Cable Dexonomy, a taxonomy designed for deformable cables, a five-fingered hand with dual symmetric thumbs and rotatable fingertips, and a demonstration-collection pipeline. It proposes decomposing long-horizon cable tasks into simple primitives and validates the approach with real-world experiments using a finite-state machine to compose tasks from primitives. The results show strong performance on short-term primitives across various cables and reveal challenges for long-horizon tasks on unseen cables, highlighting the potential for imitation learning with collected demonstrations. Overall, the work advances generalizable dexterous cable manipulation and provides a framework for dataset collection and future learning-based controllers.

Abstract

Existing research that addressed cable manipulation relied on two-fingered grippers, which make it difficult to perform similar cable manipulation tasks that humans perform. However, unlike dexterous manipulation of rigid objects, the development of dexterous cable manipulation skills in robotics remains underexplored due to the unique challenges posed by a cable's deformability and inherent uncertainty. In addition, using a dexterous hand introduces specific difficulties in tasks, such as cable grasping, pulling, and in-hand bending, for which no dedicated task definitions, benchmarks, or evaluation metrics exist. Furthermore, we observed that most existing dexterous hands are designed with structures identical to humans', typically featuring only one thumb, which often limits their effectiveness during dexterous cable manipulation. Lastly, existing non-task-specific methods did not have enough generalization ability to solve these cable manipulation tasks or are unsuitable due to the designed hardware. We have three contributions in real-world dexterous cable manipulation in the following steps: (1) We first defined and organized a set of dexterous cable manipulation tasks into a comprehensive taxonomy, covering most short-horizon action primitives and long-horizon tasks for one-handed cable manipulation. This taxonomy revealed that coordination between the thumb and the index finger is critical for cable manipulation, which decomposes long-horizon tasks into simpler primitives. (2) We designed a novel five-fingered hand with 25 degrees of freedom (DoF), featuring two symmetric thumb-index configurations and a rotatable joint on each fingertip, which enables dexterous cable manipulation. (3) We developed a demonstration collection pipeline for this non-anthropomorphic hand, which is difficult to operate by previous motion capture methods.

Figures (15)

• Figure 1: Our designed hand pulling the cable from right to left. (a)-(c): The hand performs a pre-grasp motion to position the cable correctly. (d): The left-side thumb and index finger grasp the cable. (e): The middle finger bends to hook the grasped cable. (f)-(h): The left-side thumb and index finger drag the cable to the left, and the right-side thumb and index finger hold the cable to prevent it from sliding back.
• Figure 2: Human DCM with Prehensile (or not), Motion between fingers and the cable (or not), In-hand or out-of-hand, and Support from the external contact (or not).
• Figure 3: Four types of goal configurations. The coordinate system shown at the lower right is the same for the human hand, our rendered robot hand, and our real-world robot hand.
• Figure 4: Three types of combinations of fingers. (a) TIC: the thumb and the index finger combo. (b) VMF: virtual middle fingers can be a combination of the middle finger, the ring finger, and the little finger. (c) [VMF+Palm]: using VMF and the palm for extra grasping, can be replaced by another TIC if exists.
• Figure 5: Demonstrations of cable pulling. See the appendix for other long-horizon manipulations.