For this first test, the spiral pattern did not include any path following
algorithm, but just changed the turning radius at specific points. Due
to Nomad's delays in reaching requested turning configurations, plus the
limited discrete turning radii which were allowed, the pattern did not
turn out very well.
A path following algorithm was introduced, but the correction gain
factors were apparently wrong.
In this case, Nomad did not recognize when it had reached an endpoint
-- the point where the turning radius should be increased (every half of
a circle). The robot kept driving at the same turning radius, thus making
a nearly perfect circle -- which is nice, but not what I wanted.
Nomad did better here, but the turning radius selected (out of the
discrete allowed radii) wasn't quite large enough, and the path following
algorithm was still not functioning.
This one was pretty good. I think it was only luck, though. I still
didn't have the path following algorithm working properly.
The path following algorithm was in place, but the correction gain
was not right for the speed Nomad was traveling (0.25 m/s), causing large
oscillations around the proper position. However, I think the pattern looks
pretty good.
I slowed Nomad down to 0.15 m/s, and this time the combination of correction
gain and speed was just right.
I tried it again the next week just to make sure, and yes, it still
worked.
Back to Coverage
Planning Research
 |
|
 |