Two-Dimensional Tactile Force Sensor Based on Chirped Fiber Bragg Grating Assisted by Convolutional Neural Network

Soft tactile sensors have become one of the key elements in broad applications such as intelligent robots, haptic interfaces, and wearable devices. In this article, we demonstrate a novel tactile force sensor based on chirped fiber Bragg gratings (CFBGs), which adopts a conventional neural network to retrieve the force magnitude and positions in a 2-Darea. Instead of traditional peak detection of the gratings, the full reflection spectra of the CFBGs are utilized to train the model, and a good prediction of force and 2-D positions is achieved after the network is well trained. Experimental results show that the obtained prediction accuracy of the force is similar to 99.925% with a test root-mean-square error (RMSE) of similar to 0.08 N and an RMSE of extracting 2-D position of similar to 0.01 cm. Due to the fast demodulation of one spectrum in similar to 12.88 ms, this characterization approach of the sensor allows it for real-time visualization and feedback during the applications. The promising results make it potential to be applied in automated surgical tools as well as intelligent robots or wearable devices.

Automated summary

In this article, a novel tactile force sensor based on chirped fiber Bragg gratings (CFBGs) is demonstrated, which uses a conventional neural network to detect force magnitude and 2-D positions. Experimental results show that the sensor has high prediction accuracy and position extraction precision, allowing for real-time visualization and feedback.