Page 28 - Human Inspired Dexterity in Robotic Manipulation

24 Human Inspired Dexterity in Robotic Manipulation

As explained in the previous section, the stability of grasp shown in
Fig. 2.6 is useful for estimating the subjective response, but this method
rather simplifies the mechanisms of the human hand. Fig. 2.7 also shows
contact force vectors computed from anatomically more realistic digital-
hand models including simplified muscle structures composed of the
moment arm and the maximum isometric force of each muscle. In this
method, contact forces and muscle activations are computed by solving a
quadratic programming problem; minimizing the sum of squared muscle
activations while satisfying two equality constraints and one inequality con-
straint. The equality constraints are the equilibrium between total wrench
applied to the object and contact forces, and that between contact forces
and muscle activations, while the inequality constraints ensure that no slip-
page occurs at each contact point. This method is applied to posture synthe-
sis by finding the thumb-tip position that can minimize the sum of squared
muscle activations within the range of motion for the thumb [21].
A modified version of this formulation is applied to compute the robust-
ness of grasp along specified unit wrench vectors [22]. In this case, the scale
factor for the unit wrench, contact forces, and muscle activations are com-
puted by solving linear programming problems; maximizing the scale factor
for the unit wrench while satisfying the same equality and inequality con-
straints as the previous formulation. This method is also applied to a data-
driven grasp synthesis by pruning undesirable grasps from many grasp
candidates based on the computed grasp quality.

2.5 CONCLUSION

There is a wide gap between robot hands and human hands, and it is difficult
to interchange the diverse knowledge accumulated in each research field
directly. However, as reviewed in this chapter, digital-hand technology will
be an interface that connects robot with human; motion analysis of the
human hand will help to achieve better motion synthesis and dexterous
manipulation in robot hands, while the unified formulation of grasp and
manipulation will help to understand the mechanics behind human hands
and subject responses arisen from product use.
Digital hand research is conducted in several research institutes; the
Virtual Soldier Research Program at the Iowa University, Transportation
Research Institute at the Michigan University, and Human Informatics
Research Institute at the National Institute of Advanced Industrial Science