C-R-M: pretty cool hexapod video ...

Now /that/ is how robots should behave - that is true fluid biological
immitation - im more than impressed. If it walked id be more amazed, but
it is the most impressive hex that ive seen i think.

Sheesh, between that and <http://beatbots.blip.tv/> , I'm really starting
to feel like my robots are lame because they don't track faces.

I know that OpenCV has face-tracking stuff... I just need to figure out
how to cram some hardware that can run OpenCV into something
robot-sized.  (Or, use an external computer, which is probably what the
hexapod is doing.)

Any thoughts on the smallest/cheapest computer board that could run
OpenCV?

Best,
- Joe

--
"Polywell" fusion -- an approach to nuclear fusion that might actually work.
Learn more and discuss via:  <http://www.strout.net/info/science/polywell/>

If you got a CMUcam or AVRcam or similar device, you could probably
get your bot acting very similar to this right now, cueing in on a
colored ball rather than a face. Get it to track back and forth, and
then jump backwards when the size of the object increases quickly,
etc.

True.  And with the right background and clothing, you could get it to
do that with a face too.  But that's cheating!

Best,
- Joe

--
"Polywell" fusion -- an approach to nuclear fusion that might actually work.
Learn more and discuss via:  <http://www.strout.net/info/science/polywell/>

Well, there's such a thing as the learning curve [meaning try the
simple
things FIRST as a step on the way to doing the more complicated
stuff],
plus how much time, money, and effort you really want to invest into
something that does no more than react to something that looks like
a face.

Plus, with a CMU/AVR cam, you can probably set the color tracking to
skin tone, once you get it tracking a ball. Also, after you try the
FIRST
step, you might decide the whole effort is just so much whimsy.

I guess different people find different things easy.  Over the weekend I
downloaded OpenCV and had a simple face-tracking demo up in about 20
minutes.  To interface a CMU/AVR cam would have taken me much longer, I
feel certain.

Of course there's still the issue of exactly what I'm going to do with
it, and where I'm going to stuff the computer.  But a robot that does no
more than react to faces is still far cooler (and more intelligent) than
any other robot I have a chance of building in the next few years.

And yes, of course it's mainly whimsy -- but let he whose robot actually
does something useful cast the first stone.  Personally I think giving
up on the idea of building a robot that's actually useful is a key
milestone in one's journey through hobby robotics.  On the other hand,
there's much to be said for getting beyond the typical
wander-around-trying-not-to-bump-into-things stage.

Best,
- Joe

--
"Polywell" fusion -- an approach to nuclear fusion that might actually work.
Learn more and discuss via:  <http://www.strout.net/info/science/polywell/>

[snip]


I would like to make a slight modification to what you say.
I think it is OK to give up on building a robot that does
something useful for the first few robot; however, having
something a robot that does something useful is still a
reasonable and desirable goal for the longer term.

 > On the other hand,

My $.02 (and less than 100 lines to boot ;-)

-Wayne

Yeah, having a demo program running on your PC is a little easier
than having one connected to a vid cam and running on your mini
sumo, etc.

Do you know whether the "spider" had its own visual processing on
board, or was connected to a remore PC?




As I indicated, I'll bet it would be just as impressive if the cam
caused
the bot to respond to circular moving objects. After all, spiders
don't
know about faces, they just react to moving things that might eat
them,
and vice versa.



Say what?

How about a robot that can patrol your house when you're away
from home, and send pictures of intruders directly to the internet.




Hide quoted text -

It was connected to a remote PC.


I'm in the camp that considers remote-controlled vehicles to be more,
well, RC vehicles than robots.  :)  But I do recognize that they face
many of the same engineering problems and there is a great deal of
crossover in the two hobbies.

Now, a robot that patrols your home on its own, recognizes any people it
sees, and gives you a call on your cell phone if anybody comes in that
it doesn't know -- THAT would start to border on a useful robot.  Though
only marginally so; it would be simpler and probably more cost-effective
to just put a camera in each room with a home entrance.  And most people
would say that this simpler solution is not a robot, since it has no
moving parts.

Best,
- Joe

--
"Polywell" fusion -- an approach to nuclear fusion that might actually work.
Learn more and discuss via:  <http://www.strout.net/info/science/polywell/>

Now, THAT is cheating, if you want to talk about cheating :).



I'm talking about autonomous control for the patrol bot, not remote
control.



Actually, in my opinion, a camera in every room is wasteful #1, and
extremely
intrusive of family member privacy #2. No one wants this.

We're roboticists, not home security technicians, after all. A well
designed
security robot will seek out sounds such as breaking windows,
recognize
intruders using PIR and motion sensors, and send pictures to the
internet
from anywheres in the house.

You don't immediately need facial recognition in order to build
something
useful.

My robots are VERY useful, thank you very much!
They keep my bank account from getting too big.
Keep the door from closing, etc ...
Ha! :-)

jcd

I should have tracked this info down and posted it originally - here is
their website: http://www.micromagicsystems.com/
You can find some videos of the thing walking, close-ups of design, etc.

Enjoy !