Fuzzy-neural net based control strategy for robot manipulator trajectory tracking

Abstract

A novel scheme with a neural network feed-forward controller and a fuzzy feedback controller is proposed for trajectory tracking of robot manipulators with unknown dynamic model. In the scheme, an improved BP network is used as feed-forward controller, which approximates to expected torque. The feedback controller is constructed based on T-S fuzzy model. The fuzzy rules are initialized according to the experiments of experts and experienced operators, which can provide better training data than that supplied by conventional feedback controller. Simulation results show that the presented scheme has good tracking performance and disturbance rejection ability.