The complex kinematic structure of a human thumb makes it difficult to capture and control the thumb motions. A further complication is that mapping the fingertip position alone leads to inadequate grasping postures for current robotic hands, many of which are equipped with tactile sensors on the volar side of the fingers. This paper aimed to use a data glove as the input device to teleoperate the thumb of a humanoid robotic hand. An experiment protocol was developed with only minimum hardware involved to compensate for the differences in kinematic structures between a robotic hand and a human hand. A nonlinear constrained-optimization formulation was proposed to map and calibrate the motion of a human thumb to that of a robotic thumb by minimizing the maximum errors (minimax algorithms) of fingertip position while subject to the constraint of the normals of the surfaces of the thumb and the index fingertips within a friction cone. The proposed approach could be extended to other teleoperation applications, where the master and slave devices differ in kinematic structure.
Read full abstract