Robust Anthropomorphic Robotic Manipulation through Biomimetic Distributed Compliance

Kai Junge,Josie Hughes

2024-04-14

Abstract:The impressive capabilities of humans to robustly perform manipulation relies on compliant interactions, enabled through the structure and materials spatially distributed in our hands. We propose by mimicking this distributed compliance in an anthropomorphic robotic hand, the open-loop manipulation robustness increases and observe the emergence of human-like behaviours. To achieve this, we introduce the ADAPT Hand equipped with tunable compliance throughout the skin, fingers, and the wrist. Through extensive automated pick-and-place tests, we show the grasping robustness closely mirrors an estimated geometric theoretical limit, while `stress-testing' the robot hand to perform 800+ grasps. Finally, 24 items with largely varying geometries are grasped in a constrained environment with a success rate of 93%. We demonstrate the hand-object self-organization behavior underlines this extreme robustness, where the hand automatically exhibits different grasp types depending on object geometries. Furthermore, the robot grasp type mimics a natural human grasp with a direct similarity of 68%.

Robotics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to enhance the operational robustness of humanoid robot hands in uncertain environments by imitating the structure and material distribution of human hands. Specifically, the author proposes a new type of robot hand named ADAPT Hand, which improves the robustness of open - loop operations and self - organizing behavior by introducing tunable compliance in the skin, fingers, and wrists.

### Main problems
1. **Limitations of existing robot hands**:
- Existing humanoid robot hands still have limited compliance distribution, usually concentrated in specific spatial regions.
- There is a lack of methods or indicators for systematically evaluating the incorporation of compliance.

2. **How to achieve operational robustness similar to that of humans**:
- The compliance of the human hand is distributed on different scales such as the skin, fingers, and wrists, providing the ability to make stable contact and adapt to postures.
- The paper explores whether, by imitating this distributed compliance, the robot hand can exhibit human - like behavior without complex control.

### Solutions
- **Design of ADAPT Hand**:
- **Skin compliance**: Use EcoFlex20 material to cover the contact surface to simulate the compliance of human skin.
- **Finger compliance**: The MCP joints of each finger are equipped with series - connected springs to provide compliance and allow the fingers to adaptively adjust their postures within a centimeter - level range.
- **Wrist compliance**: Achieved through an impedance - controlled robotic arm to simulate the compliance of the human wrist.

- **Experimental verification**:
- **Contact stability**: By comparing the performance of soft skin and rigid skin in sliding and rotation tasks, the advantage of soft skin in generating higher shear forces is demonstrated.
- **Posture adaptability**: Through sliding and grasping experiments, the robustness of compliant fingers in adapting to different object shapes is demonstrated.
- **Large - scale automatic grasping experiment**: Through 845 uninterrupted grasping experiments, the high success rate (93%) of ADAPT Hand in handling objects of different geometric shapes is verified.

### Conclusion
By imitating the distributed compliance of the human hand, ADAPT Hand can exhibit a high degree of robustness and self - organizing behavior under open - loop conditions, approaching or even exceeding the theoretical geometric limits. This indicates that reasonable design and distribution of compliance are the keys to improving the operational performance of humanoid robot hands.

### Formula representation
The formulas involved in the paper are mainly concentrated on force - displacement curves and force measurement results. For example:

- Force - displacement curve:
\[
F = k\cdot\Delta x
\]
where \(F\) is the force, \(k\) is the stiffness coefficient, and \(\Delta x\) is the displacement.

- Shear force measurement:
\[
F_{shear}=\mu\cdot F_{normal}
\]
where \(F_{shear}\) is the shear force, \(\mu\) is the friction coefficient, and \(F_{normal}\) is the normal force.

These formulas ensure the accuracy and repeatability of the experimental results.