PiezoSight: Coupling vision based tactile sensor neural network processing and piezoresistive stimulation for enhanced piezo-vision hybrid sensing
Vision-based tactile sensors (VBTS) have become ubiquitous in robotics for contact and touch measurements due to their minimal instrumentation costs and exceptional high-resolution feedback. However, in the literature, VBTS often suffers from the requirement of extensive calibration and is limited t...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-05-01
|
| Series: | Materials & Design |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S0264127525003193 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Vision-based tactile sensors (VBTS) have become ubiquitous in robotics for contact and touch measurements due to their minimal instrumentation costs and exceptional high-resolution feedback. However, in the literature, VBTS often suffers from the requirement of extensive calibration and is limited to a specific range of forces. Recent advances in flexible, sensitive, and robust sensors have presented high potential for excellent use cases in robotics applications. The integration of graphene in piezoresistive sensors opened new horizons for such applications. This study presents PiezoSight, an encapsulated graphene-soaked textile in a stretchable elastomeric structure integrated with a VBTS. Piezosight overcomes VBTS weaknesses so that from the piezoresistive graphene textile, PiezoSight can detect a wide range of forces for robotic perception applications, varying from tiny forces for a soft touch and slip detection starting at 0.01 N up to an elevated force of 8 N. From VBTS, PiezoSight can infer high-resolution information, such as the direction of the contact measurements. Piezosight can sustain 10000 cycles with different compression rates at a strain of 40% with sensitives at 0.09 kPa-1. A machine learning LSTM model was used to train the data for different designs for using such sensors in unsupervised environments. |
|---|---|
| ISSN: | 0264-1275 |