Position Calibration Method for Large size Industrial Robots Based on Random Forest

Most industrial robots are unsuitable for variable production systems because they are taught using the teaching playback method. In contrast, the offline teaching method has been developed, but it has not progressed because of the low positioning accuracy. Therefore, several studies have proposed methods to calibrate for positioning errors using neural networks. However, it is difficult to identify the factors of positioning errors because the structure of neural networks is not clear. Herein, we applied the random forest method, which is a type of machine learning method, and constructed a prediction model for positioning errors. A large industrial robot was used, and three-dimensional coordinates of the end-effector were obtained using a laser tracker. The model to predict the positioning error from end-effector coordinates, joint angles, and joint torques was constructed using the random forest method, and the positioning error was predicted with high accuracy. The random forest analysis demonstrated that joint 2 was the primary factor of the X. and Z-axis errors. This suggested that the air cylinder used as an auxiliary to the servo motor of joint 2 was the error factor. The positioning error norm was reduced at all points using the proposed calibration.