Decoupling Research on Flexible Tactile Sensors Interfered by White Gaussian Noise Using Improved Radical Basis Function Neural Network

Abstract

Research on tactile sensors to enhance their flexibility and ability of multi- dimensional information detection is a key issue to develop humanoid robots. In view of that the tactile sensor is often affected by noise, this paper adds different white Gaussian noises (WGN) into the ideal model of flexible tactile sensors based on conductive rubber purposely, then improves the standard radial basis function neural network (RNFNN) to deal with the noises. The modified RBFNN is applied to approximate and decouple the mapping relationship between row-column resistance with WGNs and three-dimensional deformation. Numerical experiments demonstrate that the decoupling result of the deformation for the sensor is quite good. The results show that the improved RBFNN which doesn’t rely on the mathematical model of the system has good anti-noise ability and robustness.

Keywords

• White Gaussian noise
• Decouple
• Conductive rubber
• Flexible tactile sensor
• Radical basis function neural network.