Research On Staubli Industiral Robot Calibration Algorithm And Experimental Verification

The accuracy of an industrial robot is an important indicator of the robot’s performance.Nowadays, the repeatability of the robot has reached a very high level, which can guarantee to0.1mm or even less, but the absolute accuracy of the robot is generally relative low, and thedifference between each robot is large, typically in millimeter or centimeter. With the widelyused of robot off-line programming technology, it has become an urgent problem in the fieldof robotics to improve the robot’s absolute accuracy.In this paper, by taking Staubli TX60L as a research subject, adopting kinematiccalibration method, using Leica AT901laser tracker measuring system and the parameteridentification method based on differential model, I identified the exact DH parameters of therobot. The main contents and conclusions are as follows:1) Aiming at the problem of robot calibration and combining the actual geometricalstructure of Staubli TX60L robot, every joint’s link coordinate system was established byusing classical D-H method, then the norminal DH kinematic parameters were obtained. Onthis basis, the detailed forward kinematics was derived and the solution was calculated. Then,the simulation test for the kinematic model was taken with HMI software which was writtenby Matlab. Finally, I verified that the forward kinematics algorithm written in Matlab wasconsistent with the algorithm in CS8C controller.2) In order to make the model more complete, this paper proposed that we need toconsider not only the error of the robot, but also the additional bias which is caused by themisalignment between the base frame constructed by measuring system and the real robot baseframe. Then, combining the MDH kinematic model which is suitable for robot calibration, theerror equation was derived based on differential method. Through designing a simulation testfor the error equation, the feasibility of the error model was proved.3) By using QR decomposition of the identification Jacobian matrix,28identifiableparameters were determined. Then, the optimal measuring pose set was obtained by adoptingO1to be the observability index and using DETMAX algorithm to select the optimal poses inthe candidate measuring pose set. For the identification of error parameters, a weighted Levenberg-Marquardt algorithm was used and the corresponding Matlab program is writtenaccording to the flowchart of the algorithm.4) Calibration experiment was taken by using laser tracker. By analyzing and processingthe data of measuring points, the deviation values of the norminal parameters were obtainedand compensated to the robot controller. By comparing the pose error before and aftercompensation, I found that the average position error was reduced from1.65mm to0.28mmand the average orientation error was reduced from0.692°to0.067°, so it indicated that theabsolute accuracy of the robot had been effectively improved.