Robust Tracking Control of Rigid Robotic Manipulators Based on Fuzzy Neural Network Compensator

Abstract

The problem of robust tracking control using a computed torque method and a fuzzy neural network (FNN) compensator for a rigid robotic manipulator with uncertain dynamics and external disturb signals is presented in this paper. Neural networks which have versatile features such as learning capability, nonlinear mapping and parallel processing, is difficult to obtain true teaching signals and to apply to a wide range of real-time control. However, for fuzzy control, it is not necessary to build mathematic model, and its control mechanism accords with people's logic although being short of schematism during design is its shortcoming. FNN has the advantages of both fuzzy systems and neural networks. This paper presents a novel method to obtain true teaching signals for the FNN and overcome the real-time control problems existing in the neural network control. The simulation results show that the effects of large system uncertainties can be eliminated and asymptotic convergence of the output tracking error can be guaranteed by using a FNN compensator in the closed loop feedback control system for the rigid robotic manipulator.