|
[DPRG] Borenstein's UMBMark
Subject: [DPRG] Borenstein's UMBMark
From: dpa
dpa at io.isem.smu.edu
Date: Tue May 20 17:00:52 CDT 2008
Howdy,
Hi Steve!
You wrote:
> > Or you might try different tires. jBot's are the "Road Rage"
> > after market tires:
> >
> > <http://www3.towerhobbies.com/cgi-bin/wti0001p?&I=LXAGB9>
> >
> > that gives jBot good enough odometry to run the 100 foot
> > square without compass or IMU, I bet they'd work pretty
> > well on the UNI robot as well if the tail wheel sideways
> > drag is not causing too much slippage.
> >
> I think jBot's design somewhat resists turning. So, if it's driving along,
> and you bump the back of the robot, it will still continue pretty much the
> same direction. On the other hand, a "normal" three wheeled robot will
> easily be bumped off course doing the same thing.
This is probably true, but I'm not sure it matters. The "normal"
three wheeled robot, like the contest entries of Whip by Ron Grant,
Daphne by Dale Wheat, and Erik's Teeter-Totter, can also be
easily bumper off course. However, as long as they are accurately
tracking their locations it doesn't matter. They will correct,
same as jBot, even if jBot is a little harder to perturb.
> Also, one reason for inconsistancy in turns could be because it's so easy to
> turn. Again, jBot requires a very deliberate (and consistant) amount of
> force to turn 90 degrees.
True, and the situation that Randy describes may well be because
of the scrubbing sideways friction of the tail wheel causing the
drive wheels to slip when turning, which would vary with the
surface, and probably can't be fixed by just changing tires.
>
> Not sure if that makes sense to anyone else, but it seems reasonable to me.
> :)
Seems reasonable. The main point is that the robot must track its location
continuously, and that would seem to mitigate against a technique of stopping
to take readings from time to time. Odometry is bound to drift over time,
especially in outdoor environments, of which a level asphalt parking lot
is probably as good as it gets. So ultimately some other form of correction
is needed. However, as a sequence for learning, wheel encoder based odometry
is, I think, the right starting place, and we had a lot of success with
that at the outdoor contest last month.
best,
dpa
More information about the DPRG mailing list
|
