Abstract
This paper fabricates a flexible tactile sensor for artificial skin. The skin is fabricated with polyurethane rubber to make soft channel which is filled with biocompatible aqueous solution, which can visualize the pressure distribution by electrical impedance imaging technology. This paper establishes a numerical simulation model, uses Tikhonov, Landweber, Radial Basis Function Neural Network (RBFNN) and Convolutional Neural Networks (CNN) methods to verify in noise-free and noisy comparative experiments, and conducts single-target and multi-target experiments on the fabricated flexible tactile sensor. The simulation results show that the Root Mean Square Error (RMSE) of CNN is the smallest and the Image Correlation Coefficient (ICC) of CNN is the largest, and it can accurately display the position and contour of the target. The sensor test results show that the flexible tactile sensor we prepared has a good effect in the four algorithms for pressure image reconstruction. Through CNN algorithm, RMSE can reach less than 0.1 and ICC can reach more than 0.9. Therefore, the flexible tactile sensor can intuitively display the pressure distribution, and has a good application prospect in the skin detection of flexible robot.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have