An Improvement of Trajectory Tracking Accuracy of Automatic Sewing Robot System by Variable Gain Learning Control

In the sewing factory, non-routine tasks, especially curved surface sewing of three-dimensional products are still executed manually by human workers, because it is difficult to handle the sewing parts precisely by the automatic machine. Then, we are developing an automatic sewing robot system for their sewing. We evaluated the developed robot system and confirmed that the curved surface sewing motion is executed smoothly with low feeding speed. But, the trajectory tracking accuracy becomes bad when the feeding speed is high. Then, we applied learning control method to our robot system and confirmed that the trajectory tracking accuracy is improved sufficiently by this method even the sewing parts feeding speed is equal to human workers. However, we need much time to find suitable learning gains for getting the good result. So, we propose a variable gain learning control method which finds suitable learning gains automatically based on the trajectory tracking error of the robot arm. Finally, we confirmed that the enough trajectory tracking accuracy is achieved by the proposed method without much time and effort.