Fuzzy Wavelet Neural Network Control for Pneumatic Artificial Muscle

Abstract

A 3-DOF artificial muscle is used in the fields of medical robots. To counteract the defects of its non-linearity, a FWNN (fuzzy wavelet neural network) was proposed for actuator control with the integration of fuzzy theory and WNN (wavelet neural network). Using gradient method the learning of fuzzy WNN was performed to find optimal values of the parameters of controller. Result of simulation of control system based on fuzzy WNN was compared with the simulation result of control systems based on WNN and FNN (fuzzy neural network) controller. Simulation results demonstrate FWNN controller improves the static and dynamic performance of the actuator effectively. The training of system is faster and it has better control performance than others. FWNN control is a rather ideal method for this application.