Grasping has always been a hot topic in multidisciplinary field-anatomy, neuroscience and robotics. Recent research on robotic hand grasping has concentrated on introducing the grasp mechanism and characteristics of human hand to robotic hand. This paper analyzes the problem of generating grasp configurations for given objects. A one-step grasp synthesis is applied to a 20DoF five-finger anthropomorphic hand. Experiment results demonstrate many efficient grasps are generated by the proposed grasp synthesis. By analyzing grasp configurations, it is found that the mode of the grasp is varied corresponding to the size of the object relative to the hand. Then a principle component analysis (PCA) is carried on for the generated grasp configurations. The results indicate that the first 3 principle components account for about 80% of the whole variance of the joint angles. It illustrates that the dimension of the grasp configuration space could be highly reduced in this way.
展开▼
