点群で与えられた 3D モデルに基づいた 3 次元最適把持の自動生成の研究

抄録

Autonomous acquisition of appropriate grasp position is important for stable grasps. This paper proposes an automatic generation algorithm for getting the optimum 2D grasp and the optimum 3D grasp. First, we describe a method of deriving the contact surface geometry from the point cloud. Next, 3D point cloud of the object are acquired using a depth camera. At this time, effectiveness of the optimum 2D grasp is demonstrated through grasping by the robot hand. Also, a 3D model is generated, and the point cloud for the optimum 3D grasp search is generated. Finally, effectiveness of the optimum 3D grasp is demonstrated through numerical examples.