Page 193 - Human Inspired Dexterity in Robotic Manipulation

192 Human Inspired Dexterity in Robotic Manipulation

Fig. 10.4 Graph structure used in manipulation planning.

10.3.2.1 Initial Object Placement
For the initial object placement p ini , we consider the right-hand grasp g ri and
the left-hand grasp g li ,(i ¼ 1, …, I). Let us consider the case of CG r \CP,we
consider solving the inverse kinematics of the right arm for the given com-
bination of (g ri , p ini ). If the inverse kinematics is solvable and the arm posture
is collision free, we consider applying this configuration as a node included
in the component of the right-hand grasping space at the initial placement.
The same discussion applies for CG l \CP and CG r \CG l \CP.

10.3.2.2 Intermediate Object Placement
CG r \CP and CG l \CP manifolds are structured along two foliations. So,
any path in these manifolds may be approximated by a sequence of manip-
ulation paths. We make use of the Visibility-PRM algorithm [2, 6] to cap-
ture the connected components of these manifolds. Visibility-PRM tends to
generate a minimum number of nodes covering CG r \CP and CG l \CP.
The steering method used to connect two nodes is based on the computa-
tion of Type1 paths. By using the Visibility-PRM, we can expect that the
number of regrasps included in the solution path is not large.