Page 78 - Human Inspired Dexterity in Robotic Manipulation

74 Human Inspired Dexterity in Robotic Manipulation

space. From the control-theoretical point of view, the system to be manip-
ulated is underactuated, as it has only a single control input, and producing
the prescribed rest-to-rest movements with the suppression of residual
vibrations requires a certain skill to be learned.
Bearing these considerations in mind, the experiment was conducted in
2 days. In the first day, the subjects were familiarized with the experimental
setup and learned the unusual dynamic environment. The subjects were
requested to make the reaching movements in a natural way, on their
own pace, trading off the speed (as fast as possible) and the comfort (as com-
fortable as possible). Each subject made 120 trial movements. A movement
trial was considered to be successful if at the target point the position devi-
ation was less than 0.0125 m, and the velocity deviation was less than
0.0125 m/s. These tolerances were set for the hand and each bead. The sub-
jects were provided with an audio feedback generated by the computer for
successful trials.
On the second day the experiment was slightly modified as we fixed the
movement time T ¼ 2.5 s. A trial was successful if, in addition to the posi-
tion and velocity tolerances set in the first day, the subject was able to com-
plete it within 0.3 s deviation. Again, successful trials were accompanied
by the audio feedback. For the convenience of the subjects, a time bar/sem-
aphore was displayed in continuous gradations of the blue (beginning) and
red (completing) colors. The experiment was conducted in two blocks. In
the first block (40 trials, learning phase) the subjects recovered the motor
memory formed on the first day of practice. In the second block (80 trials,
recording phase), data of the position and velocity of the hand and the sim-
ulated masses were collected for analysis. These data were recorded at the
tact frequency of 100 Hz sufficient to analyze human movements with an
average reaction time of 200 ms.

5.5.1 Hand Mass Identification
To use the minimum hand-force-change model, one needs to know the
mass of the hand. For this purpose, on the second day of the experiments
(30 min after the reaching movement experiments), the subjects moved a
single mass-damper-spring virtual object as outlined in Section 5.3. In this
experiment, the parameters of the haptic simulator were set as follows: m o ¼
1 kg, F ¼ 1.8 N, Ω ¼ π rad/s, k ¼ 600 N/m, b ¼ 0.05 Ns/m. The haptic
force along the movement line was computed by Eq. (5.40), while those in
the lateral directions were modeled by a virtual spring of stiffness 600 N/m.