Highly efficient recognition of similar objects based on ionic robotic tactile sensors

Area of Science:

• Robotics
• Materials Science
• Artificial Intelligence

Background:

• Robotic tactile sensing is crucial for object recognition and interaction.
• Current limitations in sensor performance and algorithms hinder accurate recognition of similar objects.

Purpose of the Study:

• To develop a high-performance flexible tactile sensor.
• To create an intelligent algorithm for enhanced tactile recognition.
• To improve robotic capabilities in distinguishing similar objects.

Main Methods:

• Fabrication of a flexible sensor with pyramidal microstructure and ionic gel coating.
• Development of a Pressure-Slip Dual-Branch Convolutional Neural Network (PSNet) for feature extraction and fusion.
• Tactile experiments involving object recognition, specifically on different types of leaves.

Main Results:

• The sensor demonstrated excellent signal-to-noise ratio (48 dB), low detection limit (1 Pa), high sensitivity (92.96 kPa⁻¹), fast response (55 ms), and stability over 15,000 cycles.
• The PSNet achieved a 97.16% recognition rate for leaves, outperforming human recognition (72.5%).

Conclusions:

• The developed flexible tactile sensor and PSNet offer a significant advancement in robotic tactile sensing.
• This technology holds great potential for applications in bionic robots, intelligent prostheses, and human-computer interaction.