Repetitive Motion Planning Of Redundant Manipulator Based On Recurrent Neural Network

In practice,redundant manipulators are often required to track target trajectories and perform repetitive motions in a specified area.Then,the proposed inverse kinematics model can solve this problem effectively.The gain module was traditionally set as a fixed parameter,and the motion control method could adjust the deviation of each joint angle in real time.However,the convergence speed of the error was not fast enough and had no anti-interference performance.This paper put forward a recurrent neural network algorithm based on variable coefficients,which could plan and control the repetitive motion of redundant manipulators,and then designed a motion controller.Finally,it discussed the realization of the model with noise interference,and the concrete iterative control algorithm was given.The main work of this paper was as follows:1.In view of the versatility of the robot inverse kinematics,the minimum performance index on the velocity layer was used as the quadratic function of the speed vector of the manipulator.Besides,by combining the robot inverse kinematics equation,this paper established a unified and coordinated redundancy analysis scheme.In order to further control the joint angles in real time,Lagrange multiplier was introduced to transform quadratic programming into linear matrix equation.Considering the error between the analytical solution of the model and the theoretical one,this paper designed a quadratic programming solver to transmit the results to the controller,thus driving the manipulator to move repeatedly.2.Aiming at the convex quadratic programming problem,a generalized varying-parameter recurrent neural network(GVP-RNN)control method based on time-varying coefficients was proposed,which built a fast convergent nonlinear convex quadratic programming model.In addition,to analyze the stability of the proposed algorithm,this paper used a positive-definite Lyapunov candidate function to prove that the residual error of learning algorithm converges to zero exponentially.At last,compared with the existing algorithms,the effectiveness and superiority of the proposed model were verified by theoretical analysis and simulation.3.Aiming at the repetitive motion system of redundant manipulators,a recurrent neural network based on exponential function was proposed,which called exponential varying-parameter neural network(EVPNN).For the noise disturbance existed in the realization of model,a time-varying exponential parameter was designed to accelerate the convergence speed,so as to estimate the upper bound of the residual error and control the progressive correction of the actual trajectory by the manipulator.Compared with the experimental results of existing algorithms,the effectiveness of the proposed scheme in anti-jamming was verified.