Figure 9: Average final endpoint error/feature (left), and deviation
from straight line trajectory in
feature space (right) for
a 16 tracked features and 6 controlled DOF problem
When controlling the full 6 DOF of the manipulator we do not get quite
the same robustness against model errors as in the 3 DOF case. Model
errors worse than
yield a significant and increasing proportion
of divergent trials. To get near
convergence we needed to
restrict
to be
or above. In a test of 50 trials in this
range, two did not converge. One showed oscillatory behavior, and the
other diverged immediately.
In this experiment we ran the controller to its convergence limit
rather than halting at a fixed accuracy. A remarkable
of the
trials with the best model achieve zero final visual error. This is
down to
when the model error is
. In general, average
visual endpoint errors as well as trajectory errors remain low even at
low model accuracies, as can be seen in fig. 9.