DPRG
DPRG List  



[DPRG] On contests

Subject: [DPRG] On contests
From: Chris Jang cjang at ix.netcom.com
Date: Wed May 30 21:55:20 CDT 2007

(Ed Okerson)
>Speaking for myself, I find it hard to get excited about the contests
>anymore.  The outdoor events are interesting, and I would love to compete
>in one of those, but there just isn't enough time in the day to get an
>entrant ready.

(David P. Anderson)
>> We've had this discussion on and off over the years, and the conventional
>> wisdom has always been that the membership just won't build robots without
>> contests as an incentive.
>>
>> Now I'm probably a peculiar deviant from this wisdom.  My problem has
>> usually been the opposite --- that there are no contests designed to
>> show off what my robots are built to do.

(Chris Jang)
As a new guy and oddball in the DPRG, I believe the main reason why there
are not more robots is simple economics. They are incredibly expensive to
develop. I think it is comparable to writing computer games. It takes so
much work and requires such a breadth of knowledge and experience that
even small teams fail to finish in a reasonable time frame.

Part of the expense is hardware. A capable field robot will not be cheap.
The vehicle must be rugged and somewhat powerful. It will cost hundreds
of dollars. Computers and sensors will be several times the cost of the
vehicle. This tends to push the cost up into the low thousands of dollars.

I have seen others do things inexpensively as I too have. But I think this
path leads often to a low ceiling on capability. I'm not saying it is bad
to go cheap. I think it is good. It avoids costly development and mistakes
before acquiring enough experience to design a "production" vehicle (which
will be expensive).

The other major expense is software development which can always become a
bottomless pit if it goes wrong. And most amateur builders did not study
robot programming in school. Robot software is not like any other kind of
software I have ever encountered. It becomes its own specialty rather much
like computer games. Design and programming patterns arise from the
intended application.

The last expense is theory. This is the single greatest expense for me.
After about a year of abortive software development and reading - I think
I see where image processing ends and stochastic inference begins. But I'm
only at the point of knowing - what I don't know, what I don't need to
know and what I do need to know. So I study, write code, and learn.

And if you are like me, then you had to teach yourself electronics before
even starting any of this! This took me literally 6 to 12 months before I
had even a basic intuitive grasp. It is out of the question for someone
who doesn't have much free time.

Yes, I think that if a robot is designed for a specific contest, then it
can be done more economically. But when people face the significant
investment of time and love in a hobby, they want to make something
interesting to them and special. So they then build different robots that
intrigue them. A side effect is that they cost more and may be suboptimal
for any particular contest.

jBot is a very shrewdly designed machine - 6 wheels using RC components
with odometry, GPS/IMU and SONAR for obstacle avoidance. The computer is
a microcontroller. The only thing missing (excuse me David, my recollection
of a conversation we had last year may be inaccurate) is that the Kalman
filter does not incorporate odometry readings into the process model. But
otherwise, it is pretty close to optimal selection of technologies for what
it does - much like the GPS/IMU with odometry and LIDAR for SLAM augmented
with camera video is an optimal solution for a DARPA Grand Challenge style
race (e.g. Stanley the VW Touareg).

Most people, myself included, have never progressed far enough to be so
shrewd about design. Or perhaps we have other interests. I do observe there
is often a natural convergence of design for any given set of requirements.
Given outdoor waypoint contests, most robots will be 4 to 6 wheeled or
tracked EOD (e.g. packbot) sized machines with GPS and odometry.

Anyway, that is why I think there are few robots. Building them is costly
in many ways. It requires significant investment of time, money and
education. I believe it is really much like building your own airplane or
writing a large computer game. It can take years.

More information about the DPRG mailing list