Gripping Point Determination for Bin Picking Using Heuristic Search

Random bin picking is often considered primarily to be an object localization problem. However, the determination of a suitable grasp in a given situation is an equally important part of picking chaotically stored objects for which only few solutions exist. This paper presents a novel approach by representing the localized workpieces and possible grasps by a tree and using heuristic search to determine a feasible gripping solution in a short time. The proposed heuristic function is based on collision testing and a statistical evaluation of the likelihood for a collision-free grasp. Finally, the presented search method is experimentally validated.