[Robotgroup] Drive a Droid proposal... How to detect the "fence"?

[Doug Evans]

Odometry error is generally more pronounced in the theta value (the
heading), in my experience. However, I agree that it is not the best method
for the task as described.

If it were my project, I would place some IR detectors at the front of the
robot (down low) to detect a "curb" which is painted in such a way that it
reflects the IR signal (as I believe has already been proposed by another
member). Probably IR Sharp range finders would be more than sufficient. The
robot controller would employ an obstacle avoidance behavior at a low level,
preempting the "controlled" movement.

As a fallback, it might be advisable to also employ an ultrasonic sonar
sensor at a higher vantage point to avoid bumping into anything (or anyone)
in the ring (or partially in the ring). 

As a failsafe, there needs to be a remote kill signal (aside from the remote
being utilized by the folks) so that an RG person can avoid personal injury
to spectators. This might be something as simple as a 38khz remote and a
corresponding sensor.

Sounds like a fun project, whether or not it gets funded by First Night.
Just the kind of thing that makes the kids grin.

-de




-----Original Message-----
From: robotgroup-bounces at puremagic.com
[mailto:robotgroup-bounces at puremagic.com] On Behalf Of Bryan Bishop
Sent: Tuesday, July 08, 2008 1:07 PM
To: The Robot Group Mailing List
Subject: Re: [Robotgroup] Drive a Droid proposal... How to detect the
"fence"?

On Tuesday 08 July 2008, Clendon Gibson wrote:
> If you are good with geometry, then you can put encoders on the
> wheels and calculate your position by dead reckoning.

We discussed this last time the group gathered, and it became apparent 
that the amount of loss and difference due to the wheels turning (and 
so on) is significant to the extent that the encoders will miss 
information from time to time. The buildup would be too much of an 
offset from actual data. 

Though I didn't do any calculations to see if it the buildup would be 
significant within the 30 second period that a user would be driving it 
around -- maybe after 30 seconds the buffers could be purged and it 
could return to home (center) position? 

It's not a matter of time, but rather the moves that are made with the 
spinning and the going forward and backwards, some generating more 
error than others, so it's possible to have somebody walk up and make 
it do weird things to make it damage the 'barrier' etc. before those 30 
seconds are up. 

So, I don't think dead reckoning is going to work.

- Bryan
________________________________________
http://heybryan.org/
_______________________________________________
Robotgroup mailing list
Robotgroup at puremagic.com
http://lists.puremagic.com/cgi-bin/mailman/listinfo/robotgroup