Research On Closed-loop Calibration Method Of 6R Robot Based On Machine Vision

The wide application of industrial robots including series robots and parallel robots improves the degree of industrial automation.But the absolute positioning accuracy of robots is low when they leave the factory,so they can not be controlled by off-line programming to be used to industrial production which is more complex,more flexible and require higher precision.In this paper,a new kinematic calibration method based on machine vision and point constraint with auxiliary of distance constraint is proposed to improve the absolute positioning accuracy of 6R series robots.Firstly,the modeling part uses MDH model.The inverse solution of the robot is derived,and add the judgment of joint attitude.This method only calibrates the joint parameters between the 1st joint axis and the 6th joint axis,and the other joint parameter errors are incorporated into the robot installation errors and tool errors.Secondly,in the measuring part,machine vision including high resolution camera and image recognition,high shape accuracy standard ball and gauge block are used as the main tools.The measurement includes point constraint measurement and distance constraint measurement to solve the problem of poor calibration effect of pure point constraint.The point constraint measurement takes the ball center as the fixed constraint point and takes the camera optical axis installed at the end of the robot as the constrained part.The distance constraint measurement takes the gauge block as the distance reference of the ball position.In the process of measurement,two focal lengths are used to solve the problem of poor calibration effect of the 5th and 6th joints and tools.Thirdly,the parameter identification part adds optimization solution to solve the parameter identification equation on the basis of gradient method,and uses the first order difference quotient method to calculate the Jacobian matrix in the parameter identification equation.Before the experiment,the measurement process is simulated in MATLAB,and the feasibility and effect of the calibration method are tested with parameter identification.Finally,the calibration experiment is carried out with the St?ubli TX60 robot as an example.The average value of absolute moving distance errors is decreased from 1.912 mm before calibration to 0.220 mm after calibration,while the distances range from 80 mm to 440 mm.The average repetitive positioning error of different joint posture which means the direction of shoulder,elbow and wrist,is decreased from 1.637 mm before calibration to 0.143 mm after calibration,and the error is the distance between each point and its midpoint.