Research On Grasping Detection Strategy And Grasping Path Planning In Non-Cooperative Autonomous Grasping Task

The on-orbit grasping technology of the space manipulator has been widely used in space services such as on-orbit maintenance of failed spacecraft,space assembly on orbit,space-assisted parking,and orbit cleaning.Compared with the cooperative situation,the grasping technology of non-cooperative objects has a broader application prospect.However,due to the lack of information communication,there are also greater challenges.The premise of its realization is the application of accurate and efficient grasping detection technology.At present,the artificial control method adopted by non-cooperative target capture has many disadvantages such as high task cost and low efficiency,and it is difficult to meet the requirements of busy on-orbit service tasks.Autonomous grasping technology based on path planning has the advantages of precise operation and high efficiency and can solve this problem,but there are usually problems such as discontinuous planned paths and difficulty in scheme verification.In this paper,based on the autonomous grasping of failed spacecraft,the research on the key technologies of grasping detection and grasping path planning is carried out.Aiming at the problems that the failed spacecraft cannot provide its position and orientation information and cannot directly observe the grasping parts,and the grasping platform is not easy to adjust the position,a grasping detection strategy based on a cascade network is proposed to realize the grasping detection of the failed spacecraft.Acquiring data through vision systems,the PVN3 D algorithm is used to estimate the pose of the failed spacecraft,and the pose data in the camera coordinate system is obtained.Then the pose data is converted to the coordinate system of the manipulator,and the pose transformation is performed on the solid point cloud reference model established for the spacecraft on the ground.Use the Point Net GPD algorithm with the constraints of the grasping mechanism to grasp and detect the transformed spacecraft point cloud model.Even if the camera cannot observe the part,the spacecraft point cloud model is at the same position,equal in size and complete compared with the entity.So that a suitable grasping scheme can be evaluated.Aiming at the problem of planning and optimizing the grasping path of the space manipulator,a path-planning method optimized by the improved ant colony algorithm is proposed.The vision system is used to acquire image data,and the image data is rasterized to construct the octree map.The ant colony algorithm is used to complete the planning of the crawl path on the octree map.Various B-spline curves are used to smooth and optimize the motion path.The optimized paths are evaluated comprehend-sively,and a quintic B-spline curve is chosen to optimize the motion path.The D-H parameters are used in the analysis of the manipulator to clarify the law of the manipulator’s motion and provide a theoretical basis for the half-physical simulation experiment.Aiming at the difficulty of scheme verification,a hardware-in-the-loop simulation experiment is designed.In the ROS system,the URDF model of the gripping mechanism of the manipulator is established,and Move It is used to perform collision detection on the model.The Gazebo simulation environment is built to verify the communication of the simulation,the execution status of the action of the mani-pulator grasping mechanism,and the situation of the camera capturing images.The internal and external parameters of the visual system of the physical simulation experiment platform are calibrated.In Gazebo,the scene consistent with the physical simulation environment is built to verify the feasibility of the path planned by the ant colony algorithm,and the physical manipulator is controlled to verify the feasibility of the path planned by the ant colony algorithm.The simulation results show that the vision system can extract the object field information more accurately,that the capture detection strategy based on the cascade network can realize the pose estimation and capture detection of the failed spacecraft,and that the ant colony algorithm can plan a suitable avoidance Obstacle paths,and that a quintic B-spline curve can optimize smooth motion paths.