PiezoTac: Fingerprint-Inspired Piezoelectric Tactile Sensing for Enhanced Texture Discrimination

The tactile perception capability of the robotic fingertips is crucial for achieving dexterous operation and effective human-machine interaction. Utilizing bionic technology to develop anthropomorphic tactile sensing in robotic fingertips represents a promising approach. However, the development of biomimetic tactile sensing based on the segmented curve layout of fingertip fingerprints remains in its early stages. This study proposes a fingerprint-inspired piezoelectric tactile sensor (Piezo-Tac) featuring eight piezoelectric films arranged in a segmented curve layout. A multi-channel wireless acquisition circuit was designed, achieving a sampling rate of 1 kHz. The sensor responds to frequencies exceeding 200 Hz, aligning with the Pacinian corpuscle's frequency range. Comparative experiments showed that the curved layout outperforms a straight-line layout in directional signal response. Additionally, a horizontal sliding experimental paradigm was designed, and a Support Vector Machine (SVM) was employed for two tasks: direction angle recognition and texture surface discrimination. With a training set ratio of 0.3, the sensor achieved recognition accuracies of 96.19% and 98.10% for direction angle and texture surface classification, respectively. While independent time-domain features yielded satisfactory results, fused time-frequency features provided slight improvements. This work advances biomimetic tactile sensing by exploring the segmented layout of fingerprint curves, offering new insights for robotic tactile sensor development.