Massively parallel in-sensor skinomorphic computing

Area of Science:

• Materials Science
• Robotics
• Electronics Engineering

Background:

• Traditional tactile sensing separates sensors and processors, hindering real-time processing for skinomorphic electronics.
• This limitation poses a significant challenge for the advancement of intelligent robotics and wearable technology.

Purpose of the Study:

• To propose and demonstrate a massively parallel in-sensor skinomorphic computing scheme for tactile perception.
• To overcome the limitations of traditional tactile sensing by enabling parallel sensing and processing directly within the sensor.

Main Methods:

• Fabrication of a 32x32 flexible capacitive pressure sensor array with high uniformity and endurance.
• Integration of the sensor array with a memristive crossbar array for in-sensor computing.
• Experimental demonstration of parallel sensing, restoration of pressure patterns, and feature extraction.

Main Results:

• Successfully demonstrated parallel sensing and restoration of broken pressure patterns (e.g., 'NJU').
• Achieved direct, parallel extraction of textural features from complex pressure patterns using networked sensor and memristive arrays.
• Showcased tactile information compression through parallel feature extraction.

Conclusions:

• The proposed in-sensor computing scheme enables real-time, high-throughput tactile perception for intelligent skins.
• This approach overcomes the physical separation bottleneck in traditional tactile sensing.
• Opens new possibilities for advanced skinomorphic electronics in robotics and wearables.