Research On Grasp Of Aerial Manipulator Based On Three-dimensional Information Of The Object

The unmanned aerial vehicle(UAV)has been widely used in many fields such as aerial photography,power inspection,address survey,and agriculture.The UAV equipped with a robotic arm,which called aerial manipulator,improving the active operation capability of the UAV,and the grasp task can be completed by the aerial manipulator.In order to improve the success rate of object grasp operation of the aerial manipulator,the grasp estimation method based on the three-dimensional information of the object is studied.The main contents of this paper are as follows:Firstly,the hardware system of the aerial manipulator is built,and the kinematic model and dynamic model of the aerial manipulator are established respectively.According to the constructed system model,a fuzzy sliding mode attitude controller is constructed.Considering the disturbance caused by the motion of the robotic arm during the operation task,the auto disturbances rejection control technology is combined with the fuzzy sliding mode controller.And the disturbance that the UAV received is estimated and eliminated to improve the robustness of the aerial manipulator.Secondly,the principle of visual ranging methods is analyzed and the depth camera is selected.The coordinate system of the grasp scene is designed and the relationship between the object coordinate system and the world coordinate system is derived.According to the depth camera model and the internal parameter matrix of the depth camera obtained through camera calibration,the relationship between the depth pixel in the depth image and the coordinates of the point in the real-world space is derived,and the 3D point cloud information calculation is realized by the relationship.Thirdly,the grasp estimation method based on the three-dimensional information of the object is studied.A YOLO based object region extraction method of the depth image is designed.The YOLOv4-tiny detection model is used to detect the object in the depth image,and the object region in the depth image is extracted by the background subtraction.Taking the 3D point cloud information of the object as the input,realizing the object grasp estimation through 6-DOF GraspNet.In view of the characteristics of the aerial manipulator during the operation,the angle weight is introduced to the grasp evaluation function.Meanwhile,the distance between the grasp and the plane is as the constraint to optimal the grasp,improving the success rate of the grasp operation.Finally,the performance of the designed controller and the effectiveness of the designed grasp estimation method are verified through experiments.Simulation experiment results on the MATLAB show that the designed fuzzy sliding mode controller based on auto disturbance rejection technology has good anti-disturbance performance.In the real-world experiment,the results of the aerial manipulator grasp experiment show that the grasp estimation results based on the optimized 6-DOF GraspNet can be applied to the real-world grasp operation of the aerial manipulator,and the success rate is higher than that of the ARG-GRCNN.