Mapping human hand motion to dexterous robotic hand

Abstract

Grasp planning is one of key issues for robotic dexterous hands to accomplish the desired tasks; this paper describes the implementation of grasp planning for robotic dexterous hands. Human grasp strategy is learned through master-slave manipulation, Human hand and robot hand operate as the master and the slave, respectively. A method of fingertip mapping is developed based on virtual fingers in Cartesian space. Using the data glove as human machine interface, joint angles are recorded and digit trajectories are computed using forward kinematics. The kinematics of movement of all digits is analyzed during grasp. In order to give an evaluation of the mapping algorithm we quantify the motion mapping results by using a general distance as a characterization of comparability between human hand and robot hand posture. Our method considers the grasping posture between all the fingers not only the fingertip positions. The fingertip trajectories during grasp are recorded in Cartesian space. A virtual environment is developed to visualize the motion of the human hand and the robot hands. Visualization of the 3D workspace is accomplished. The mapping results are verified in the virtual environment, illustrate the validity of the mapping algorithm.