The aerial manipulator is a flight robot system composed of a unmanned aerial vehicle (UAV) and a manipulator. The system combines the characteristics of maneuvering flight, hovering of the UAV and flexible operation of the manipulator, which can complete the operation task in flight state. At present, the researches on the aerial manipulator system mainly focus on the design and control of the system. There are a few researches on the grasp pose generation in the object manipulation task of the aerial manipulator. In this paper, a grasp pose estimation method for aerial manipulator is proposed. Firstly, a 3D point cloud extraction method which combines object detection and background subtraction is designed. Secondly, the 6-DOF GraspNet is utilized to generate grasp for the object. In order to generate executable grasp for the aerial manipulator, the grasp evaluation strategy of 6-DOF GraspNet is improved. The experimental results show that the proposed method can estimate an effective grasp pose for the aerial manipulator. The success rate of grasping for the bowl shaped object is 90%, and the success rate of grasp for the bottle shaped object is 83.3%.
Grasp Pose Estimation for Aerial Manipulator Based on Improved 6-DOF GraspNet
2022-08-19
1137121 byte
Conference paper
Electronic Resource
English