Description
This paper presents a universal method for improving the path accuracy in varying, highly dynamic pick and place processes. The proposed method is based on an iterative learning control (ILC) and is valid for serial as well as parallel robots. It extends the traditional ILC approach for purely repetitive tasks by evaluating similarities between the occurring movements. Based on this, the correction term for following trajectories can be calculated and a significant improvement in the accuracy can be obtained even for previously unknown (unlearned) motions. The performance of the method is studied and verified using an exemplary four degrees of freedom delta robot. It is shown that the presented approach improves the path accuracy up to 86% even if the occurring pick and place trajectories vary with respect to start and end position. It also outperforms conventional computed torque control methods by up to 50%.
