Automatic Grasp Planning Using Depth Image for Robotic Manipulator

Abstract

In this paper, we present a new solution for automatic grasp, which uses the depth image to reconstruct the real world at first. Then our system will recognize the object automatically by using a variant of HFC algorithm. Since the segmentation algorithm is based on primitive shape, we can use it to decide the grasp pose. In the end, our system will implement the grasp process automatically. A set of primitive shape, such as cylinder, plane, sphere, will be given. And we can generate the grasp rules for each shape in the set of primitive shape. In the end of this paper, experimental result will be given. The innovation of this paper is that we use the depth sensor to reconstruct the real world in high-precision, and because of the precision of the virtual scene we can implement the grasp process automatically. We believe that automatic grasp planning system using depth image will be an excellence solution for automatic grasp.