|
[DPRG] Re: DPRGlist Digest, Vol 48, Issue 19
Subject: [DPRG] Re: DPRGlist Digest, Vol 48, Issue 19
From: Doug Emes
kyojin at gmail.com
Date: Mon May 19 11:37:28 CDT 2008
> From: "Randy M. Dumse" <rmd at newmicros.com>
> Subject: RE: [DPRG] Tarzan's lessons learned at the DPRG Outdoor Challenge
[snip]
> So I like the idea of keeping the compass tucked down, and
> occassionally sensing where to stop to hoist it into clean air.
>
> I don't know, it's more moving parts, but then, I think we are
> perhaps more fascinated by robots with more moving parts, et?
Even R2-D2 on occasion had to "up his periscope" to figure out where
he was (dagobah bog, tattoine dunes, etc).
I am still in the infancy stages of robot building, but keeping a "low
profile" during run times would seem to be more efficient, if only for
avoiding obstacles (park benches, vines, etc). We walk upright, but
can bend at the waist / duck / even bear crawl if necessary. I don't
imagine making a wheeled robot to the same, but maybe hoisting the
compass is something to do IF the data sets show it reduces error
versus having these sensors mounted on the top deck of the robot.
(I've edited this post twice now to remove pirates / jolly rogers
references, arrgh!)
Whether done via a linear extension (akin to a car antenna) or on a
pivot arm, providing a "higher vantage point" for some sensors can be
done. I guess the need would be to test this type of configuration
and see if it provides a more accurate measurement. Something Eric
Sumner said at a recent RBNO stuck in my head (regarding the "right"
or "correct" method)
"...the engineering answer is 'is it accurate enough for the specified
needs' .."
Or something close to that (Eric, would you mind quoting yourself =)
ex: If a robots drivetrain's slip is over an inch, having millimeter
positioning isn't necessarily
better than half-inch resolution. Alexander's wheel encoder was white
tape on the actual
tread, providing one click per wheel revolution, and thats probably
"good" enough for almost
any DPRG style contest (except maybe for obstacle avoidance scenarios
if you are trying
to map onto some quadrant plane).
More information about the DPRG mailing list
|