Research On The Control Methods Of Manipulator Trajectory Tracking Based On RBF Neural Network

Industrial robot( manipulator) occupies a very i mportant position in the worl d industrial production, it is always a hot topic in the control of joint trajectory tracking. But the manipulator is a m ulti input multi output system, it has the characteristics of nonlinear and strong coupling, so it is a big c hallenge to develop high precision, high speed tracking control algorithm.Radial basis function(RBF) neural network has strong nonlinear mapping ability, its structure feature, learning algorithm and application in manipulator control are analyzed in detail in this paper, and th e theory is verif ied about the nonlinear approximation characteristic of RBF neural network.The basic structure of the manipulator system is analyzed in detail, the dynamics equation of two DOF serial m anipulator is derived by using the La grange Euler method, the inertia characteristics, Coriolis force and centrifugal force characteristic, gravity torque characteristic are analyzed. The manipulator nonlinear system model is established based on S function in th e MATLAB, and the dyna mics model is transformed into two order dif ferential equation form, so that the design of t he algorithm introduced handily.If the manipulator system is not the presence of external disturbances, and gravity term of dynamics equation is precisely kno wn, the traditional PD control can obtain ideal tracking effect. But this system is not existence with no external force and no external interference. Accord ing to this problem, a new point to point control of manipulator with interference and gravity uncertainty is proposed. On the Basis of PD control structure, the RBF neural network is introduce d to realize the gravi ty compensation, and a robustness analysis is also given with respect to t he approximation error of RBF neur al network. The Lyapunov theory is used for analyzing the stability of the system. A simulation study on a two-joint rigid arm reveals that the ideal trajectory can be obtained in tracking accuracy, convergence speed and anti-interference with the new scheme.The advantages and di sadvantages of sliding mode varia ble structure control is analyzed about the manipulator system in this paper, it is shows that the output torque of the manipulator can appear chattering cau sed by single sliding mode control, it will lead to the error in the trajectory tracking. According to this problem, a new adaptive sliding mode controller based on radial basis function neural network m ethod is proposed in the m anipulator trajectory tracking. The sliding m ode variable structure control is used for resisting interference and guaranteeing the system stability, the RBF neural network is introduced to overcom e the system uncertainty so as to reduce the switching gain through self-learning ability. The input of RBF neural network is the sliding function, and its output is the switchi ng gain which can be adjus ted adaptively. Under the condition of existing model erro r and external i nterference, The MATLAB simulation study on a two degrees of freedom and thre e degrees of freedom r igid manipulators reveal that the good perfor mance can be obtained both in tracking the trajectory and reducing the chattering.