Computing Convex Grasping Positions for Parallel Jaw Grippers with an Integrated Boundary Layer Mesher

The automated handling of workpieces is one of the main enablers for automatic production systems and requires a stable grasping point determination. To select suitable grasp candidates for STL based models with a planar parallel jaw gripper we propose a two stage algorithm. First, the STL file is remeshed to create a boundary layer containing triangles for every surface of the handling object through a shifting of its edges. In the second step, antipodal grasping positions are determined with the use of local convex hulls, which are generated via a Delaunay3D triangulation, and through an orientation examination to classify the area as convex or concave. Based on this classification all convex points are cross checked to identify pairs of antipodal Points and the associated grasping position. An evaluation of our approach was performed on a test set containing 10 different industrial and everyday objects ranging from gears and shafts to sockets and a freeform body. For all objects a highly diverse solution set could be generated and through the remeshing process the number of grasping positions could be enlarged by at least 38% for all objects.