Study On Serial Robot's Calibration Technique

The absolute positioning accuracy and repeatability of positioning accuracy of the robot is a measure of the error in the teaching robot,error repeat positioning accuracy of the robot as the main index to measure the value of the system,and in the robot off-line programming system,the absolute positioning accuracy as the main index to measure system error.However,the industrial robot has higher accuracy and repetition the absolute accuracy is low,it is difficult to meet the accuracy of off-line programming work,the urgent need for the robot kinematic calibration to improve the absolute accuracy.The Kawasaki RA010 N robot as the object,using the theory of rigid space position has completed the conversion of the robot body in the same coordinate system and between different coordinate systems,based on the application of matrix form to determine the position and attitude of robot end effector in the base coordinate system,and establish the differential equations of motion of robot according to the differential movement,mapping analysis of end effector differential movement and joint motion.The kinematic modeling of robot using DH model,solved the robot mechanism of pose matrix in robot base coordinates.And the singularity of the DH model is improved and obtained the MDH model,and establishes the relationship of robot positioning error and robot kinematics parameters the error between.Combining with the positive kinematics model of robot,the robot inverse solution using neural network.To solve the robot as training samples,training by double genetic neural network,the input and output function of network is Tansig and purelin function,realize the nonlinear mapping from the robot work variable space to joint variable space,so as to avoid the complex formula is derived and calculated.The combination of visual measurement and calibration technology of the robot,namely eye off hand,using the method of binocular vision for robot calibration experiments,and according to the robot target dot recognition algorithm is analyzed,through the data processing software to obtain the calibration results,the mean distance error decreased from 0.27 mm to 0.06 mm before calibration using the kinematic parameter calibration results,compensation,confirm the feasibility of the method.