Two sets of experiments (Fitts’s targeting and pouring
tasks) were conducted to evaluate manipulation performance
through robotic arm control instead of computer simulation of
3D graphical model (used widely for input modality testing).
The robotic arm manipulation was selected because it was task
centered and can provide more thorough subject assessment of
input modality usability and performance. Further, robotic arm
control allows subjects to easily view arm movement in three
dimensions. The focus of this study was to enable individuals
with SCIs at the most common mid-cervical neurologic levels,
which lead to gross motor function of the shoulders and elbow
flexion, to effectively operate a device in 3D. Two subjects
with Cervical-4/5 and Cervical-6 SCIs and three subjects
without disabilities (ages 25-42) were recruited for these
experiments. For the two subjects with SCIs, one had limited
wrist function while the other did not and used wrist braces for
fixating the wrist. Neither subject had hand or finger
movement. During the experiments, each device was
positioned according to subject preferences for robotic control.
During the Fitts’s targeting task, the relative accuracies of each
of the input control modalities were compared by the subjects’
abilities to touch the tip of a pen held by the robotic arm
(Figure 4, left side) to different sized and positioned targets as
quickly as possible. The two targets, sizes 2.8 x 2.8 cm and 8.0
x 8.0 cm, were alternately placed at locations 40cm and 70cm
from the base of the robotic arm. Fitts’s Law result for this
targeting task performed five each is shown in Figure 5. The
slopes for the 3D joystick, traditional joystick, discrete
(keyboard), continuous (keyboard) and hybrid (keyboard)
control modalities were 2.7, 3.08, 3.79, 4.73, and 3.58
respectively. The 3D joystick had the smallest slope or the
highest index of performance (IP) (reciprocal of the slope). A higher IP for the 3D joystick indicated a greater human rate of
information processing during the targeting task.