The picking of unsorted parts from boxes and other containers is called Box Picking – the supreme discipline of robotics. One or more 3D cameras first generate a 3D point cloud of the crate contents. Complex algorithms are used to calculate the position and orientation of the workpieces in the crate. In addition, there are also collision avoidance algorithms that determine the best possible robot path for gripping a workpiece. Newer solutions implement neural networks and deep learning to make the mentioned algorithms more flexible and robust.
We have integrated various 3D machine vision systems so that the user can freely choose the solution that best suits the application. As an example, the video shows a bin-picking application with a KUKA robot, a Roboception 3D camera and a Robotiq gripper system.