Page 19 - Human Inspired Dexterity in Robotic Manipulation
P. 19

Digital Hand: Interface Between the Robot Hand and Human Hand  15


              2.2.2 Link and Joint of the Digital Hand
              The digital-hand model consists of a surface mesh, explained in the next
              section, for visualizing the surface geometry and a link structure for control-
              ling the posture and movement of the hand. As we reviewed in the previous
              section, the control of hand motion by the musculoskeletal structure is com-
              plicated, because the movement of one specific joint is produced by more
              than one muscle, and one specific muscle is contributing to the movements
              of several joints.
                 To simplify the control of hand motion, the link structure of the digital-
              hand model has fewer degrees of freedom (DoF) than those of the human
              hand. Fig. 2.2 shows examples of the typical link structure of digital-hand
              models. The range of motion is small for the second to fifth CMC joints,
              and movement of these joints are often ignored (Fig. 2.2, left) [8]. However,
              if the movement of these joints is completely ignored, hand posture such as a
              cupped palm cannot be reproduced. To solve this problem, other digital-
              hand models place two joints at the second to fifth CMC joints (Fig. 2.2,
              middle, and right) [9,10]. Usually, IP joints are described by one (flexion-
              extension) DoF hinge joint, while MCP joints of the second to fifth fingers
              are described by two (flexion-extension and adduction-abduction) or three
              (flexion-extension, adduction-abduction, and pronation-supination) DoF
              ball joints. CMC joints and wrist joints are also described by two or three
              DoF joints.
                 Joint centers of the hand are determined from a two-dimensional ana-
              tomical image, three-dimensional medical images, or motion capture
              (MoCap) data of the hand. The anatomical image based methods utilizes
              landmarks and/or hand dimensions obtained from the two-dimensional
              image, the X-ray image, or palmar-side image of the hand, for estimating






                                DIP                   DIP                  DIP
                               PIP                  PIP                   PIP
               IP                   IP                    IP
                             MCP                   MCP                  MCP
              MCP                   MCP                  MCP

                 CMC      CMC     1    CMC       CMC   1     CMC       CMC   1
                                  2                    2                     3
              (A)                   (B)                  (C)                 6
              Fig. 2.2 Link structures of the three different digital-hand models. (A) No CMC DoF [8];
              (B) Santos hand [9]; (C) DhaibaHand [10].
   14   15   16   17   18   19   20   21   22   23   24