Page 83 - Human Inspired Dexterity in Robotic Manipulation

Modeling and Human Performance in Manipulating Parallel Flexible Objects 79

Table 5.3 VAF percentages in matching of the experimental data with the minimum
hand-jerk (left) and minimum hand-force-change (right) models
Hand Bead Bead Hand Bead Bead
Subject (%) 1 (%) 2 (%) Subject (%) 1 (%) 2 (%)
Subject 1 93.3 87.1 82.1 Subject 1 93.5 88.4 82.8
Subject 2 76.6 94.3 81.2 Subject 2 77.0 94.5 85.7
Subject 3 95.1 82.8 87.8 Subject 3 93.5 83.5 85.0
Subject 4 95.9 90.1 90.4 Subject 4 95.5 93.0 90.2

∗
where v k and v are the mean values of, respectively, v k (t) and v*(t) on the
movement interval.The VAFpercentages,calculatedforall thesuccessfultri-
als for all the subjects, are summarized in Tables 5.3. From the comparison of
thesedata,oneconcludesthatforthegivenreachingtaskandtheparametersof
the flexible object, the two theoretical models are practically identical.

5.6 DISCUSSION

Here, we summarize additional issues that are relevant to the results reported
in this chapter, and that are a part of the ongoing research work.
1. Kinematics vs. dynamics. Whether the CNS plans movement in kinematic
or dynamic variables is an old issue in the computational neuroscience [9,
25, 26]. A straightforward interpretation of the minimum hand-jerk
model suggests that the CNS may plan and execute arm movements
in the task space of hand coordinates, taking into account the dynamic
properties of the novel environment the arm is “embedded” in. Indeed,
the optimality criterion (5.1) is purely kinematic. It is assumed that the
hand can be driven by the CNS ideally as the hand position is treated as
the control input. Under such an assumption, hand dynamics do not play
an independent role in establishing an optimal solution. They are used
for the estimation of the driving force when the optimal solution is
found.
If trajectories of the human hand in reaching movements can be pre-
dicted without taking into account the inertial properties of the arm, it
would imply that the arm dynamics are already “prewired” in the CNS
while the dynamics of the novel environment, represented by the flex-
ible object, are acquired by learning. Otherwise, one may resort to the
task-space counterpart of the minimum hand-jerk model, the minimum
driving hand-force-change model which does take into account the
hand dynamics.