With the development of the intelligent industry,the industrial robot takes more parts in manufacturing activities as well as the applications are shifting from the teach function based to intelligent applications.Nowadays,even the repeatable positioning accuracy of industrial robot is very low which is below 0.1mm,the absolute positioning accuracy is much too high which can be more than 1 mm and even 2-3mm.The low absolute positioning accuracy becomes the barrier of the development of intelligent industry.This thesis addresses the research and experiment to enhance the absolute positioning accuracy based on the kinematic model and function of robot and the physical truth of the usage of industrial robot in the industrial field.The DH model of the robot KUKA KR500-3 is established and the closed form solution of inverse kinematic function is deduced in this thesis based on the theory of the kinematics of industrial robot and the character of the structure of the KR500.A compensation method-which is called compensator-to improve the absolute positioning accuracy is addressed based on the physical truth that the model and parameters stored in the robot controller can not reflect the real parameters of the robot.Two directions of information including transferring commands to the robot and receiving the data from the robot can be compensated to enhance the positioning accuracy of robot.The tool coordinate calibration algorithm is developed,which can use the precise model to complete computation to get the tool coordinate in precise model to meet the need of the high accuracy application of industrial robot.For the condition that the geometric factors influence the error of robot the most,this thesis identifies the errors of parameters to calibrate the robot by an optimized algorithm using least square method and furthermore to establish the precise model.Addressed an optimized algorithm of inverse solution of kinematic based on calibrated parameters,which only cost low resource and time.The accuracy can satisfy the need of industry tasks.The result of validation of the real task of robot shows the accuracy is enhanced.As the application of industrial robot is complicated and the condition that the geometric method model cannot handle,this thesis addresses a method that trains two ANNs from the sampled data in two different directions to fit two functions to represent the solutions of the forward and inverse kinematic functions of robot to fit the compensator.The accuracy of robot positioning is determined by the training error of the LM-BP ANN.The validation in one working project shows that the ANN method can correct the robot to be more accurate. |