Hierarchical Tactile Sensation Integration from Prosthetic Fingertips Enables Multi-Texture Surface Recognition

Area of Science:

• Robotics and Artificial Intelligence
• Materials Science and Engineering
• Biomedical Engineering

Background:

• Advanced prosthetic hands require sophisticated tactile feedback for improved dexterity and control.
• Current prosthetic technologies often lack the nuanced sensory capabilities of human touch.

Purpose of the Study:

• To design and integrate highly stretchable liquid metal tactile sensors (LMS) into prosthetic fingertips.
• To evaluate the ability of LMS to distinguish surface textures and sliding speeds.
• To demonstrate hierarchical tactile sensor integration for complex surface recognition.

Main Methods:

• Photolithography was used to manufacture highly stretchable liquid metal tactile sensors (LMS).
• LMS were incorporated into prosthetic hand fingertips for tactile sensing experiments.
• Time-frequency features from LMS signals were extracted to train machine learning algorithms (KNN, SVM, RF, NN).
• Performance was evaluated for single-finger speed/texture detection and multi-finger complex surface classification.

Main Results:

• Individual LMS successfully distinguished between different speeds of sliding contact and surface textures.
• Hierarchical integration of four LMS signals enabled the distinction of ten complex multi-textured surfaces.
• The neural network (NN) algorithm achieved the highest accuracy (99.2 ± 0.8%) in multi-textured surface classification.

Conclusions:

• Highly stretchable liquid metal tactile sensors offer a promising solution for advanced prosthetic hand control.
• The demonstrated hierarchical integration of multiple sensors significantly enhances the sensory capabilities of artificial hands.
• This technology could lead to more intelligent and responsive prosthetic devices, improving user experience and functionality.