Alt-BEAM Archive
Message #00628
To: beam@corp.sgi.com
From: Steven Bolt sbolt@xs4all.nl
Date: Fri, 19 Feb 1999 11:07:22 +0100 (CET)
Subject: [alt-beam] Re: hmmm...i wonder which subject
On Thu, 18 Feb 1999, Terry Newton wrote:
> >I'm aware of Terry's work, which is very interesting. However, as
> >you say, he didn't seem to gain much performance by adding a uC to
> >a Nervous Net - though it increased complexity quite a bit.
>
> Nevertheless the walker is much more functional with the
> processor than without. Not for intelligence reasons but for
> the on/off timer, wakeup on approach, a monitor that resets the
> microcore if it saturates, sensor processing, reverse timing, etc.
> If I would have hardwired all those neat functions it would of been
> much more complicated.
But it seems that it would be less complicated and just as
functional if you let the uC control the motors with nothing but
dumb drivers in between. From your documentation I got the strong
impression that the Nervous Net was a problem rather than a
solution in your case:
A big problem with controlling microcore walking robots is the
control signals don't cause the same motion every time,
---8<---
Trying to make it do specific things is frustrating, rather you
have to let it find a solution that it likes. If you don't like
what it's doing, you have to find learning rules that trick it
into behaving the way you want it to,
What I was looking for in this discusion is an example of synergy,
a combination of uC and Nervous stuff which is more capable then
either would be on its own. In your pages, it is easy to find
arguments for having a uC in control, but the Nervous Net family of
things doesn't appear to hack it. You conclude that
I'm left with the feeling that much more will be needed to make
these devices useful in the real world.
And proceed with a (very interesting :) project based on a
programmable device, with nothing Nervous on the parts list.
Which brings me to a next subject:
[ Is the `brain' overrated? ]
I have great respect for Terry's work on machine learning. And that
bridge will no doubt have to be crossed, but imho much more than
the present level of capability can be reached without any learning
at all.
It is said that an ant follows something like a dozen fixed rules.
Put it in a simple environment - like a Robot Jurassic Parc - and
what it does looks no more intelligent, capable or complex than the
behaviour of today's robots. But look at what happens when you put
that ant in the `real' world! All of a sudden its actions acquire
purpose and meaning. Now where did that come from?
It's not supplied by the ant's brain. Purpose and meaning happen
because the ant and its environment evolved together, into a single
complex entity, which doesn't work when you take it apart.
And to make it work, the ant needs plenty of sensor bandwith and
mechanical ability, since that's what connects it to its
environment.
Brain power is not the weakest point of our `bots - it's their
strongest point, the part least in need of further development.
Their greatest deficiency is sensors, with mechanical ability a
close second. Like most mobile life forms, an ant has a good sense
of smell and it marks its environment using pheromones. Our bots
lack that entire dimension. An ant senses touch, force of touch,
temperature - the list goes on and on. And it senses with every
part of its body.
The ant's mechanical agility, its power to weight ratio, its speed
and fuel efficiency - all make a joke of what we call a mobile `bot.
Imho the way forward for mobile robots with an interesting level of
autonomy is development
- Downward from beter and more sensors,
- Upward from better bodies,
With not that much more brain in between.
To me, Mark T's "Walkman" is interesting because every actuator is
also a sensor. It has fairly impressive mechanical abilities, and
its brainpower is not overdone compared to its other specs.
Interaction with its environment is what makes it smarter than
you'd believe from merely studying its schematic.
Best,
Steve
----------------------------------------------------------------------
# sbolt@xs4all.nl # Steven Bolt # popular science monthly KIJK #
----------------------------------------------------------------------
------------------------------------------------------------------------
eGroup home: http://www.eGroups.com/list/alt-beam
Free Web-based e-mail groups by eGroups.com
Home