Trajectory Tracking Control Of Manipulator Based On Gaussian Process Regression

The manipulator is currently the most widely used automated mechanical device in the field of robots and related technologies.Since the advent in the last century,manipulator has been developing rapidly at home and abroad.The manipulator itself can be improved in structure,and different drives,sensors and actuators can be selected.In addition,with the advancement of control science,except traditional control methods,new control algorithms are also emerging.Therefore,the trajectory tracking control of the manipulator has more attempts in the selection of control strategies.In this paper,the Gaussian process regression algorithm is used to study the system identification and trajectory tracking control of the robotic arm system.The main research contents of this paper are as follows:Firstly,the Lagrange system is modeled on the manipulator and then a Gaussian process regression algorithm applied to trajectory tracking control of the manipulator is proposed.The algorithm is based on the traditional calculated torque control of the manipulator.A control law is developed using feedforward compensation based on Gaussian process regression.The Gaussian process is used to learn the unknown system dynamics of the manipulator from the training data,the mean prediction of the Gaussian process regression is used to compensate for the unknown dynamics,and the variance is used to adjust the gains.This part also performs explicit tracking error calculations to ensure that the tracking error has an upper limit.Finally,the effectiveness of the manipulator trajectory tracking control method is verified by MATLAB simulation experiments.Secondly,a sparse online Gaussian process regression algorithm considering input noise is proposed for the input noise of the manipulator system on the basis of the previous part.In this part,moment matching is used to solve the non-Gaussian distribution problem caused by stochastic test points,and the mean and variance are obtained.Then consider multiple stochastic trial points.Next,in order to solve the problem of input noise,we choose to achieve control by sparse online Gaussian process regression.The sparse online Gaussian process regression algorithm is used to solve the input noise problem.Taylor approximation method is used to give the approximate value of the posterior distribution of the measured input points,and then a sparse online Gaussian process regression algorithm considering input noise is proposed.This method can be used to process stochastic measurement points.In addition,the approximation method of multi-output function is also studied and its posterior distribution is determined.Making full use of the possibility of all sparse online Gaussian process regression algorithms considering input noise based on the above conclusions,this paper establish a sparse online Gaussian process regression algorithm considering input noise applied to the tracking control of the manipulator system.Finally,MATLAB simulation results verify the effectiveness of the control algorithm.