Page 411 - Sensing, Intelligence, Motion : How Robots and Humans Move in an Unstructured World

386 HUMAN PERFORMANCE IN MOTION PLANNING

the tasks given to subjects in this study are quite close to teleoperation tasks, the
reported results should be taken seriously by designers of teleoperated systems.
There was one exception from this pattern: The training factor had a slight
positive effect on the subjects’ overall performance in Experiment Two (see the
box “Experiment Two and combined data,” Table 7.21; the extent of improvement
is only 3.67%). The meaning of this exception is not clear. Given that the effect
disappears for the combined Experiments One + Two data (further in the same
box in Table 7.21), the small effect of the training factor for the Experiment Two
data might be an artifact due to the insufficient data or measurement errors. Or,
training may indeed improve—though only a little bit—human performance in
motion planning tasks such as ours.

Effects of the Motion Direction Factor. This factor has two components, left-to-
right direction and right-to-left direction of motion. These two tasks took place in
the same scene and with the same two-link arm manipulator. The only difference
was that in the first task one was asked to move the arm from position S (start)
to position T (target) (Figure 7.5, Section 7.2.2), and in the second task one
would go from T to S. In this study the motion direction factor happened to
have a significant effect on the subjects’ performance. In fact, the effect has
been stronger than other effects observed. Hence the motion direction factor was
included in the study, to help assess the effect of the task difficulty on one’s
performance, with or without other factors involved.
Using the same scene and the same arm in both tasks has an added advantage
that the perceived difficulty of one task over the other is then known to be “in
one’s head only.” After all, a human subject could in principle produce exactly
the same path in both tasks, which is what a robot algorithm would do. 12 The
unequal difficulty of the two tasks as perceived by the subjects is very interesting.
It suggests that human performance is limited by human motion planning skills
no less than by the task’s objective complexity. As a minimum, it demonstrates
a profound qualitative difference between the human and robot algorithms.
Why do human subjects perceive the above two tasks as completely differ-
ent? It is as if changing the direction of motion to right-to-left produces some
additional, if unclear, difficulties; perhaps it adds more possibilities for motion
planning or more ways to make mistakes. In Section 7.2.3 we made an attempt
to speculate about the reasons affecting human performance in these two tasks
(see design comment No. 4 and Figure 7.9).

12 Formally, this is not exactly so. For example, in the example in Figure 3.5, Section 3.3.2, the path
shown from point S to point T is produced using the local direction “left.” If the same algorithm
(here Bug2) now starts from T toward S, using the same local direction, the resulting path will be
different from the one shown: It will be complementary to the shown path in that it will pass around
parts of obstacles that were not passed when moving from S to T . The same is true for the arm
manipulator algorithms discussed in Chapters 5 and 6. The nature of this difference is, however,
not the same as in human performance. By simply switching the algorithm’s local direction to its
opposite, we will obtain a path identical to the one shown in Figure 3.5. Whatever rules guide human
motion planning strategies, they must be very different.