Re: The trouble with hobbyist-level robot hardware

I think its not so much that new tech isnt available or involve a cost
parameter as
perhaps the level of creative integration. The actual I/O devices old
or new simply
end up doing the same applications over and over. Newcomers often
follow the crowd
(published books on the subject)or try and utilize the plug and play
methodology.
I work "out of the box" on new ways for sensor implementation most
builders might
think unlikely, as well as actuator/motor control devices. I also build
from scratch in
both circuitry and mechanical part fabrication, both unavailable to
many builders
newcomer or not.
It stands to reason you work with what you have of can afford to buy.
Also, I really
dont think information is really shared much between builders. We all
dont write books
or publish what we learn and use, and many just rely on written
resources of those who
do/have. Lots of people here ask questions, or propose new ideas, but
rarely do you
see a thread offer start to finish information on a unique
process.(unfortunate!)

Mark

Joe Strout wrote:
In article <U2aFg.10382$kO3.883@xxxxxxxxxxxxxxxxxxxxxxxxxx>,
John Nagle <nagle@xxxxxxxxxxx> wrote:

There really hasn't been as much progress in hobbyist-level
hardware in the past twenty years as there should have been.
Computing has advanced enormously, yet most hobbyist robots are
still using Polaroid sonars, analog R/C servos, and
Basic STAMPs from the 1980s.

Here's a perspective from a newbie (well, really more of a wanna-be) to
the hobby, FWIW.

First, controllers: I've seen a lot of alternatives to the Basic STAMPs
coming out lately. Pololu has their great Orangutan boards, LEGO has
the new NXT, other people are using OOPICs or whatnot. I also carry
around a lot of computing power in the form of my PalmPilot -- I know a
few people have made these into removable robot controllers, though I
haven't seen as much of this as I would expect. Also, there is the XBC,
which is a really amazing controller.

Almost all of this is too expensive for me, though -- if it's not under
$100, I can't get it past the budget committee (i.e. my wife) right now.
Perhaps this is useful data -- there are certainly price-sensitive
hobbyists like me, and much of the current hardware doesn't sell to
them. Conversely, there are hobbyists who are largely price-insensitive
(I know several in my local club), and for that market, lowering the
prices just means leaving money on the table.

At the same time, the folks who are price-insensitive are also often
those with the least ability to put together their own PC boards, power
supplies, and other such hackery required by many of the cheaper
devices. So robotics suppliers are in the awkward position that, to
reach the (probably much larger) casual robotics market, they have to
simultaneously sell a more-complete device and sell it for less money.

First, R/C servos. They're still used mostly as output-only
devices. Although it's possible to tap into them and get
out position and torque information, this is rarely done.

I didn't know it was possible at all, except for the new digital servos,
which are way too expensive. Would you have any references on how to do
this? Sounds pretty neat.

That's something to
work on - find an R/C servo that can be reprogrammed into
communicating bidirectionally in some intelligent way.

Indeed. Wouldn't that basically turn an ordinary R/C servo into a
digital servo, of the sort used by the Robo-One-style robots?

Sonars. We have enough compute power today to do serious
sonar processing, like submarines do. But mostly, people
are still just pinging with 30 degree beam widths, just like
the 1980s. That's disappointing. We should have bat sonar
by now.

That's interesting. I'm currently working on a laptop bot, which will
have plenty of processing power, as you say. I'm also a software
engineer, and probably up to the task of writing code for any algorithm
I understand. But I don't know what algorithms you'd use to extract
more data from a sonar echo -- again, references would be appreciated.

Motor controllers with positional feedback generally are
still too expensive, as are motors with encoders.

I think back-EMF is pretty cheap, though to do it at high speed may
require a FPGA, like the XBC. But I too have been surprised at how hard
it is to find decent position encoders for a reasonable price, given
that it seems like such a simple problem.

FWIW, the new NXT motors have built-in encoders with 1-degree
resolution, and cost about $18. However, I don't know how you would
control one of these except from the NXT controller.

Acroname has some cheap encoders, but they're not
well integrated with motors.

Well, that's not necessarily a bad thing -- it lets you mix and match
your motors and your encoders, even getting motors from nontraditional
sources like junked cars or whatever. I do agree that integrated
encoders would be a lot easier, though.

We do have cheap cameras, and more is being done with them,
so there is progress. But you still can't buy Aibo-grade
components off the shelf.

What's the difference between the Aibo components and standard cameras?
I'm planning to put a USB camera on my bot (maybe two, someday), and see
what I can do with it. Vision is another thing that's probably better
done with some pre-processing in a FPGA (again like XBC), but I'll have
to make do with software. I expect to be able to do color tracking, and
maybe motion detection. If I expend a lot of effort, I may be able to
get it to find people and faces in its view. Anything else you think I
should be shooting for?

Best,
- Joe

.

Relevant Pages