关键词: 柔性触觉传感器, 解耦, RBF神经网络, 高斯白噪声

Abstract: In view of the high nonlinearity and difficulty of decoupling process of the flexible tactile sensor, an effective method is proposed to simulate the situation of a flexible tactile sensor interfered by noises in practical application. Firstly, different white Gaussian noises are added into the ideal tactile sensor model, and its mathematical model is established. Then, the K-means and recursive least squares methods are used to optimize the Radial Basis Function Neural Network（RBFNN）. The optimized RBF neural network algorithm is used to approximate the high dimensional nonlinear mapping relationship between the sensor resistance interfered by noise and deformation. The three-dimensional deformation of the sensor is decoupled by the row-column resistance based on different spreads. The decoupling results show that the RBFNN with strong robustness and anti-noise ability has good performance in approximate the highly nonlinear relationship between the sensor variables interfered by noises.

Key words: flexible tactile sensor, decoupling, RBF neural network, white Gaussian noise