DPRG List  

[DPRG] Robot in the woods

Subject: [DPRG] Robot in the woods
From: David P. Anderson dpa at io.isem.smu.edu
Date: Fri Jul 15 17:36:28 CDT 2005


Stopped by the tent-camping park at Garza-Little Elm Resorvir on my way
in to work Friday morning and tested the 6-wheel robot in the woods.

The robot was set to navigate through the woods to a point 200 feet away,
then return, and repeat that 4 times.

This is the longest run to date, 1600 feet, about 1/3 mile. and it had a total
accumulated position error of about 20 feet, or 1.25% total error.  This is
using the new odometry+inertial measurement navigation (no GPS), but with no
side-slip corrections.

Almost all of the error was in the X axis (lateral) with almost no error
in the Y axis.  That is, the robot returned almost exactly to the "starting
line" but offset sideways from the actual "starting point."

The largest error was in the first leg, about 8 feet, with each leg thereafter
having LESS error than the previous, on the order of 4 feet.  I have seen this
before, where the accuracy of the navigation appears to increase when the robot
is in motion.  Need to think about why this should be the case.

Anyway, here's a video of the first leg of the test.  I scratched a big "X" in
the ground with the heel of my shoe where the robot began, in order to have a
point of comparison for the accumulated error.  Robot does an especially nice
job of maneuvering through the picinc table/BBQ pit area on the first leg (10 Meg):


Here's all four legs of the test, (50 Meg):



More information about the DPRG mailing list