Research On Motion Accuracy Of Flexible Joint Of Robot Based On BP Neural Network

Most of the traditional industrial robots are made of rigid materials,and their load and weight are relatively large,which will produce large energy consumption in the process of operation;when the related tasks of human-computer cooperation are carried out,the larger joint stiffness has a certain operational risk for the user.Therefore,lightweight has become the development trend of industrial robots,and flexible joint robots with low weight / load ratio,high flexibility,high precision and high robustness have become the development trend of applications in various industries.However,due to its own internal elastic structure will also bring about a series of problems such as poor control accuracy,poor robustness and so on,so it is necessary to perform high performance on flexible joint robot the stability and accuracy of the control task have a bad effect.in order to enable flexible joint robots to achieve better performance,it is crucial to study flexible joints.In this paper,the flexible joint of lightweight cooperative robot is taken as the research object,and the dynamic modeling and high precision and high robustness control strategy are studied.The main work and innovation points are as follows:(1)The flexible joint of the six-degree-of-freedom cooperative robot is taken as the research object of the system,and the flexible joint with load is equivalent to a two-rigid spring system.The control strategy of feedforward and feedback is used to control the robot.Based on the analysis of the influencing factors of its structure and control precision,the joint motion simulation analysis and experimental verification are carried out,which shows that the proposed theory can effectively improve the robot precision.The dynamic model of flexible joint robot is established and the control principle of flexible joint robot is studied.(2)Due to the existence of elastic mechanism,friction and fixed servo parameters of the flexible joint,it is difficult to meet the accuracy requirements of the operation,resulting in low motion accuracy,difficulty in establishing an accurate dynamic model of the controlled object,and poor parameter tuning and poor performance during the dynamic model.To solve such problems,the BP neural network control algorithm is introduced into the PID controller to complete the design of the closed-loop feedback BP neural network control system for the flexible joint position,and the method is simulated and analyzed,showing the correctness of the proposed theory.(3)There are some defects in BP neural network control algorithm,such as slow convergence speed and easy to fall into local minima.In this paper,the parameters of neural network PID controller are adjusted online by combining additional momentum term with adaptive learning rate.According to the control characteristics of PID and the dynamic model of the controlled object,feedforward control is introduced into the speed loop and current loop of PID control.The dynamic response characteristics and control accuracy of servo motor are improved,the self-learning time of neural network is shortened,and the following error can be kept in good condition in the learning process of BP neural network.