Object Size based Fingertip Workspace Processing for Acceleration of Grasp Pose Generation

抄録

Point clouds describing fingertip workspace can be used to generate precision grasp poses with no constraints on the robotic hand structure. As during generation, the k-d tree for neighbour points searching between object cloud and workspace clouds is frequently used, and also because the O(logn) averaged searching time complexity is highly based on the number of points inside the cloud, how to reduce the number of points inside each workspace cloud by removing invalid points should be one way for the acceleration of grasp pose generation. In this paper, we propose a method to process workspace clouds based on the analysis of target grasp object and also the middle points of fingertips contacting. Experiments have been done in conditions of parallel gripper, 3-fingered and 5-fingered robotic hands.