|
[DPRG] Robot building survey
Subject: [DPRG] Robot building survey
From: Chris Jang
cjang at ix.netcom.com
Date: Mon Dec 11 21:46:55 CST 2006
> * Why aren't you building a robot?
I am writing software for my robot now.
> * Why aren't you making any progress on your robot?
I hope I am making progress. I'll really know in a few months if I can
match the performance of published computer vision research.
> * If you are making progress on your robot, why haven't you told the
> rest of us about it?
I am grateful to this list for the forum. Had people not warned me
(unanimously!) off of C++, I'd be writing performance critical code in a
language and programming style that does not optimize very well.
> * What can the club do to make building a robot more enjoyable for you?
The club has done as much as I would expect. Thanks. The biggest thing
I've learned is that odometry and an IMU, in that order, should be
considered first in a robot design. Naturally, my robot has neither. But
that's my fault.
What I hope to offer the club is reusable computer vision software
technology so we can see some amazingly cool robots running around and
solving problems.
Why?
I've figured out that I really don't enjoy trying to do everything. It's
too hard. Fabrication, electronics (digital, analogue, mixed-signal),
mechatronics, powertrain, vehicle platform, firmware, software, control
theory, signal processing, computer vision, path planning, artificial
intelligence ... it's too much! I'll be mediocre at everything, even if I
can learn how to do it.
So I decided to focus on one thing that can give value to many (as well as
to me - I need this for my robot). A few years from now, it will be
normal. But right now thousands of man hours are required to develop even
a rudimentary computer vision system from scratch. It would be crazy for
every robot using optical cameras to require the same investment. We need
reusable components.
I feel that robot building is back where personal computers were in the
1970s. The DPRG is like a Homebrew Computer Club before mass marketed
desktop computers. The barrier to entry is pretty steep. You have to
design and build the hardware. Then ... you have to write all the
software! Yikes!
More information about the DPRG mailing list
|
