Robot error compensation based on incremental extreme learning machines and an improved sparrow search algorithm

It is essential to improve the absolute position accuracy of industrial robot milling systems. In this paper, a method based on an incremental extreme learning machine model (IELM) is proposed to improve the positioning accuracy of the robot. An extreme learning machine optimized by the improved sparrow search algorithm (ISSA) to predict the positioning errors of an industrial robot. The predicted errors are used to achieve compensation for the target points in the robot's workspace. The IELM model has good fitting and predictive power and can be fine-tuned by adding fewer samples. Combined with an offline feed-forward compensation method, the solution was validated on the milling industrial robot KUKA KR160. The method has been validated on a KUKA KR160 industrial robot, and experimental results show that after compensation; the absolute positioning error of the milling robot is improved by 86%, from 1.074 to 0.154 mm. After fine-tuning the industrial robot's error prediction model using a small number of measurement points once the robot had moved to a new machining position, experimental results showed that the average absolute positioning error of the robot's end-effector was reduced by 70.76%, from 1.71 before compensation to 0.5 mm after compensation.