DPRG
DPRG List  



[DPRG] Re: Sony QRIO sensor info

Subject: [DPRG] Re: Sony QRIO sensor info
From: David P. Anderson dpa at io.isem.smu.edu
Date: Sat Dec 20 11:24:01 CST 2003

Howdy

We have these discussion from time to time on the list about
how far away we are from robotic/human equvalence, and they
bring out interesting points but ultimately always end up as
"religious" dicussions, i.e., I "believe" that humanoid robots
are just around the corner (always) vs. I "believe" that we will
not reach that point (ever).

So everyone gets to air their view of the future and then we all get
back to robots, which is what I'd like to do now.

My robots all keep track of their position using wheel odometry, which
works suprisingly well, even for the LegoBot.  However, the errors in
position, and especially in rotation (theta) are unbounded and accumulate.

The SR04 robot addresses this problem by periodically using its two SONAR
to square up with a wall, thereby resetting both theta and X or Y, depending
on which wall. 

My experience is that the theta or rotation error is the real source of
problem, especially when the robot strikes an obstacle or drives over
an obstruction.

J.Borenstein of U of Mich has suggested using a gyroscope to check for and
correct errors in theta that arise from such bumps and shocks.  His method
which he calls "gyrodometry"  is to calculate the next incremental position
>from the current position using both wheel odometry and a gyroscope.  As long
as the two are roughly the same, go with the wheel odometry.  But when the two
are markedly different, as when the robot experiences as shock or bump, go
with the gyroscope.    Here's his webpage:

http://www-personal.engin.umich.edu/~johannb/gyrodom.htm

This is what got me thinking that a full 6 Degrees of Freedom Inertial Measurement
Unit might be similarly useful when combined with standard wheel odometry.

Conceiveably, such a system would allow the robot to maintain its location even
when the wheels have lost contact with the floor, as when a human picks the robot
up and moves it to another position.

At this point, nothing I've read about the amazingly cool Sony QRIO suggestes that
the robots are doing any sort of cognitive navigation in the videos we've watched,
other than running a pre-programmed sequence which was tweaked by hand.

The GPS thing sounds interesting.  Anybody got one we can play with...???

merry christmas,
dpa




More information about the DPRG mailing list