CHAPTER 3. Visual grasp algorithm
d
f
Figure 3.9: Contour of neighbor points of the current target grasp point and
possible direction of movements of the associated finger.
· f
ci
is zero when the fingers are far from each others and from the palm,
while its magnitude is increased when a safety distance is violated.
Moreover, the barrier forces can be also employed to cope with environmental
constraints, e.g. object ground plane or other surrounding objects.
Evaluation of the quality of the grasp
The force f
i
is projected onto the tangential plane to the current reconstruction
surface at the contact point i, determining the direction of motion for the i-th
contact point:
f
i
= f
i
- (f
T
i
v
i
)v
i
,
where v
i
is the unit normal vector to the reconstruction surface at the point p
i
.
The direction of f
i
individuates one of the points of the surface close to the
current one, as shown in Figure 3.9, and it is employed by the planner to produce
the floating motion of the finger. When ||f
i
|| is higher than a given threshold
f
,
the current grasp configuration changes according to the directions of f
i
. The
choice of
f
means that forces f
i
whose norm is under this threshold can be
neglected, and when this happens for all the contact points, then the reached
configuration is the (local) optimal grasp for the current iteration. Obviously
f
affects both the accuracy of the grasp solution and the computational time,
determining the number of iterations required to converge to the (local) optimum,
and thus it must be suitably tuned considering this trade-off.
36