BP and RBF neural network in decoupling research on flexible tactile sensors

This paper proposes two decoupling methods for a flexible tactile sensor, improved back propagation neural network (BPNN) and radical basis function neural network (RBFNN). 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. Information of the tactile sensor array at different scales is also used to construct the BPNN. RBFNN is applied to approach the nonlinear relationship between the deformation and the three-dimensional force of the tactile sensor numerical model built through finite element analysis. The decoupling results show that the RBFNN with high nonlinear approximation ability has good performance in decoupling three-dimensional force and satisfies both the decoupling accuracy and real-time requirements of the tactile sensor. Different white Gaussian noises (WGN) are added into the ideal model of the flexible tactile sensors. Then the modified RBFNN is applied to approximate and decouple the mapping relationship between row-column resistance with WGNs and the three-dimensional deformation. Numerical experiments demonstrate that the improved RBFNN doesn´t rely on the mathematical model of the system and has good anti-noise ability and robustness.