Industrial Robot Geometric Parameters Calibration Based On Vision Measurement

Robot positioning accuracy is needed when it is used in automobile,electronic board assembling,medical treatment assisting.Robot positioning accuracy includes repeated positioning accuracy and absolute positioning accuracy.The former generally needs to be improved.The common way is robot kinematic parameter calibration.In the past study,the laser tracker,which experimental process is complex but high accuracy,and the least square estimation are regard as the regular method.It is expensive to experiment.To reduce calibration cost,keep measurement process simply,effectively and improve accuracy in case of measurement disturbances.This paper aims to mainly improve two aspects.First,the improvement of least square method by the anti-disturbance pose selection base on the camera measurement.Second,the improvement of accuracy by the improved artificial fish swarm algorithm.The detail works as follow.The robot kinematic model is modeled by DH,and the error modeling base on link differential transformation is re-derived.The function relationship between robot endpoint errors and kinematic parameter errors is established,which is a linear expression of error model.IRB120 is the object of this paper,the simulation for DH and error model are carried out to verified.The Zhang's method is used to camera calibration based on the high accuracy board.Then the Shah' s way is to calibrate the transformation relationship among the robot arms,camera,calibration board and the base coordinate.Their error analysis also show that they can achieve the experimental expectation.Then the identification based on the error model is studied,the redundant parameters are eliminated and the minimum pose number and anti-interference pose are selected by Jacob matrix,observability index and max O2 methods.For higher accuracy in identification,this paper uses the Levenberg–Marquardt algorithm(LM)and Extend Kalman Filter(EKF)to solve it,but the codes show that these algorithm will encounter the singularity in their iteration.To solve the singularity of identification meanwhile yield the high accuracy,this paper proposes to identify the parameters by the improved step adaptive artificial fish swarm algorithm,and the object function is modified to suit for the algorithm.Finally,a vision measurement software,which control the PC and robot and measure the robot endpoints error mainly,is developed.The experimental data are measured by it.The results show that the method in this article improve the accuracy to 50.13%.