DPRG
DPRG List  



[DPRG] DPRGlist Digest, Vol 48, Issue 19

Subject: [DPRG] DPRGlist Digest, Vol 48, Issue 19
From: dpa dpa at io.isem.smu.edu
Date: Tue May 20 12:42:59 CDT 2008

Howdy

Randy wrote:
> I imagine only using this when precision turns, like might be
> needed in dpa's Outdoor Challenge for the Borenstein test, where
> odometry on asphalt just can't give that great an accuracy. I
> imagine stopping at the end destination, raising the compass,
> making the turn, and lowering it again before striking out on
> another leg.
>

Randy, the Borenstein squares do not require precision 90 degree
turns, but rather accurate tracking of the robot's position
thoughout the run.  Erik's robot was able to get back within a
few feet of the target with odometry alone this way.  John did
it with an encoder and a compass.  jBot can do the same with its
IMU switched off.  None perform accurate 90 degree turns.  They
are not necessary.

The phrase "before striking out on another leg" suggests that
the robot only needs accurate heading information while turning
and not while attempting to drive generally.  Not so.

SR04 is able to pause from time to time and realign it's odometry
in the DPRG CanCan contest because the size and shape of the
course is known, all walls are parallel and perpendicular, and the
robot can measure the distance and angle to those with its sonar.
So it can correct not only its error in theta but also in X and Y.
But we don't have the luxury of a known environment with off-road
robots.

So letting the odometry drift for a while and then "pausing to
get our bearings" by taking compass readings can correct for errors
in HEADING, but not for accumulated errors in LOCATION.  It's too
late for that.  The IMU (or compass or whatever) must be read
continuously for the location to be calculated accurately.  What
looks to the casual observer like the robot "driving a straight
line" in fact consists of tiny steering corrections made many
times per second.

The Borenstein square is not an end in itself, but rather a method
for calibrating the robot's instantaneous calculation of its own
location -- as will be required when navigating freely around
obstacles, for example, in the next challenges.  Concentrating
on precision turns as the solution to this challenge will lead
away from the more general case rather than toward it.

Hope this helps.
best regards,
dpa


More information about the DPRG mailing list

Copyright © 1984 - 2006 Dallas Personal Robotics Group. All rights reserved.
Website Design by NCC

For the latest robot news visit robots.net