The Study of Decoupling Methods for a Novel Tactile Sensor Based on BP Neural Network

This paper proposes a decoupling method for a novel tactile sensor based on improved Back Propagation Neural Network (BPNN). In the numerical experiments, the number of hidden layer nodes of the BPNN is optimized and k-fold-cross-validation (k-CV) method is also applied to construct the dataset. Furthermore, information of the tactile sensor array at different scales is used to construct the BPNN, which enhances the performance greatly. Numerical simulations show that the BPNN with strong nonlinear approximation ability plays an important role in decoupling mapping relationship between resistance and deformation of the tactile sensor, which significantly increases the decoupling accuracy and satisfies the real-time requirements of the multi-dimensional tactile sensor array. Copyright © 2013 IFSA.