Inverse Kinematics of High Dimensional Robotic Arm-Hand Systems for Precision Grasping

Abstract

Conventionally, the arm and the hand are considered separately for robotic manipulation and grasp. The hand configuration and the wrist pose are obtained during grasp planning and afterwards, the arm configuration is found to accommodate the wrist pose. However, without considering the arm during grasp planning, this approach can become very inefficient as significant time and computation power are spent on evaluating and planning grasps that are simply unreachable. This paper presents an efficient method for obtaining a desired configuration of the arm and the hand simultaneously for precision grasping. Our method assumes that contact points and contact normals are given for the desired grasp. Inspired by the human grasp strategy, our approach implements a thumb-first strategy to narrow down the search space and increase the success rate. To precisely fulfill all fingers’ requirements, the inverse kinematics (IK) solution for the position and orientation of each finger are regarded as independent tasks. These tasks are organized in a well-designed hierarchy and the resulting joint movements from each level are combined through nullspace projection with a nullspace enlargement method to ensure the correctness of the results. Comprehensive simulation results will demonstrate that the proposed method can significantly improve the performance of classic IK algorithms.