In-container Robot Palletizing with 3D vision-based Datum Point Positioning

Palletizing inside the container faces inefficiency of manual palletizing and positioning inprecision of robot. To solve these problems, it is necessary to consider two important procedures: robot palletizing scheme analysis and datum point positioning inside the container. A 3D vision-based positioning algorithm, which was combined with robot kinematics was proposed and then executed. A D-H table was built for robot RS050N to achieve the homogeneous transformation matrix; then, whole corner positioning algorithm based on 3D vision was designed, which included segmentation, down sampling, filter and an algorithm designed for boundary fitting in point cloud. It can reduce data volume, improve efficiency, and remove the point cloud which interferes positioning precision, such as stiffeners and bumps at corners. In the experiment, this algorithm can realize precise positioning for industrial robot palletizing in the container, meeting MAE of 1.715 mm and RMSE of 1.976 mm.