Spatially distributed biomimetic compliance enables robust anthropomorphic robotic manipulation

Area of Science:

• Robotics
• Biomimicry
• Mechanical Engineering

Background:

• Human manipulation relies on distributed hand compliance for robust interaction.
• Current robotic hands often lack the inherent compliance for human-like dexterity.
• Spatially distributed compliance is key to advanced manipulation capabilities.

Purpose of the Study:

• To investigate if mimicking human hand's distributed compliance in an anthropomorphic robotic hand improves open-loop manipulation.
• To introduce and evaluate the ADAPT Hand, a novel robotic hand with configurable compliant elements.
• To demonstrate human-like grasping behaviors and robustness in robotic manipulation.

Main Methods:

• Designed and built the ADAPT Hand with configurable compliant elements on skin, fingers, and wrist.
• Quantified the impact of compliance on individual components compared to rigid configurations.
• Conducted automated pick-and-place tests and object grasping experiments with diverse geometries.

Main Results:

• The ADAPT Hand demonstrated grasping robustness near the geometric theoretical limit across over 800 grasps.
• Achieved a 93% success rate in grasping 24 diverse objects in a constrained environment.
• Exhibited passive adaptation leading to hand-object self-organization and varied grasp types.

Conclusions:

• Spatially distributed compliance significantly enhances robotic hand manipulation robustness and enables human-like behaviors.
• The ADAPT Hand's design effectively leverages passive adaptation for successful object manipulation.
• The hand's grasp strategies showed a 68% similarity to natural human grasps, validating the biomimetic approach.