Gripping Point Determination and Collision Prevention in a Bin- Picking application

Bin-Picking is a complex subject. Attempts of industrial realizations are often too slow or too unreliable. Many approaches strongly focus on detecting the pose of objects inside the bin. While this is an essential part of bin-picking, industrial realizations also need to have a robust strategy to pick the detected objects out of the bin, even in complex situations. We address this issue by separating the object pose detection from the task of finding an appropriate gripping position. Our goal is to find a suitable position for the gripper and avoid collisions with the bin or other objects. This collision-free and fast gripping point determination will be presented in this paper. The basic idea is to generate several potential gripping configurations and to rate them, primarily based on the probability of collisions with any obstacles inside the sensor point cloud. The implemented approach has produced good results in experiments and industrial applications.