Research On Grasp Planning Of Five-fingered Dexterous Hand For Space On-orbit Maintenance

China will build long-term running space station. In order to accomplish on-orbit maintenance of space station in the dangerous environment, it needs to develop the robot with five-fingered hand and the abilities of object recognition and autonomous grasp planning scheme. This kind of intelligent robot cooperation with astronauts will have great benefits. The dexterous hand is considered as the reference model which has been independently researched and developed, and a visual sensor is utilized to acquire data from environment, including depth information. So the relevant basic research will be implemented, and the contents are as follows:(1) The depth data are used in the open source library PCL to build the 3D digital point cloud model of the environment. Based on the theories in PCL, such as the VFH descriptor and region growing segmentation, an algorithm of object recognition in random environment is proposed, and the experimental verification is carried out.(2) The complete 3D point cloud model of the target object is mapped out by the SampleConsensus-Initial-Alignment algorithm. According to the force-closure grasping rules, the grasping algorithms of two fingers and three fingers are proposed to find the corresponding grasping configuration, namely the grasping vector and the grasping points. So far, the grasping parameters related to the target object have been extracted.(3) The D-H parameter method is used to establish the kinematics model of the dexterous hand. The principle of virtual work is used to establish the static model of the finger. And the dynamic model is established by Lagrange formulation. The control problems of the hand are solved, such as how to get the desired position, generate the desired torque and speed.(4) The grasping parameters of the object and the model of the hand's kinematics, statics and dynamics are integrated to establish the virtual system in the OPEN INVENTOR(OIV), including the dexterous hand and the target object. An algorithm is proposed to calculate the virtual force between the tip of the finger and the object. The conditions of the stability grasp are improved, so that it is applicable to virtual grasp. Finally, the simulation of the grasping experiment is implemented in the virtual system to verify the correctness of the algorithm.Through the above research, the recognition and localization of the target object in the environment are realized. The mapping from partial view to the complete 3D point cloud model of the target object and the extraction of the grasping parameters are completed. Using the kinematic and dynamic modelof the dexterous hand in the virtual system achieves the goal that the imported target object is successfully grasped by the hand.