Alt-BEAM Archive
Message #03758
To: Dave Hrynkiw dave@solarbotics.com
From: Steven Bolt sbolt@xs4all.nl
Date: Wed, 26 May 1999 11:58:02 +0200 (CEST)
Subject: [alt-beam] Re: Stryder / Cog-like machine in 5 yrs WAS [non-tech / LONG]
On Wed, 26 May 1999, Dave Hrynkiw wrote:
> >Ummm... I find Stryder very interesting, but the practical
> >application isn't obvious to me.
>
> Proof of concepts don't really have to be practical, do they?
No, but you were putting it up for hospital duty:
> My opinion is that it's a very viable layout for larger machines
> designed to work in linoleum-floored buildings, like a hospital.
> My impression is that Stryder is one of Mark's experiments at combining
> heads with bodies. The fact that it uses such an interesting walking gait
> has merit in itself.
Of course. It's a great `bot!
> > If a `bot won't lift its legs, why not use wheels instead?
>
> But it DOES lift it's legs, but just not that far. There's arguments for
> interpreting action (ie: you know which way it's going to go just by the
> position it's legs are in), low disturbance of environment (even a few cm
> in a garden will disturb lots less soil than wheels). Etc, etc.
I rather doubt that for instance a walking tractor would disturb
less soil than a wheeled one. It might have more traction, perhaps,
and might step over certain obstacles that would stop even the
large 4wd variety I once worked with. But frankly, I'd feel more
comfortable defending the art and fun aspects of walking devices.
> > > But what I will comment on is, THAT TO ME, Mark's work is
> > > parallel to Brooks' work a decade ago with Ghengis and the like,
> > > but using a different approach. I fully expect Mark to be working
> > > on a Cog-like machine in the next 5 years too.
> >
> >No disrespect intended! I like Tilden robotics a lot. But would you
> >perhaps like to take a bet on that? Cog-level complexity/computation
> >using Nervous Nets in five years? I'll take a good Canadian whisky,
> >you can have any equivalent European bottle :)
>
> I'll take that bet!
Great! That puts the decisive date at May 26, 2004. I may remind
you every May 26 until then :)
> But no others, ok? If I lose, I don't want it to be the reason
> Solarbotics goes out of business! But I want a more exact
> criteria of performance. Head/hand coordination, like "I see it,
> I grab at it" like Cog had on "Robots Rising"?
You may not be improving your chance of winning here, because
Mark's work may take him elsewhere. But I have no objections. There
is a paper about that aspect of Cog called:
"Self-Taught Visually-Guided Pointing for a Humanoid Robot" by
Matthew Marjanovic, Brian Scassellati and Matthew Williamson. They
implemented the "visuomotor coordination" system which allows Cog
to "saccade its pair of two degree-of-freedom eyes to foveate on a
target, and then maneuver its six degree-of-freedom compliant arm
to point at that target." They summarize "This task requires
systems for learning to saccade to visual targets, generating
smooth arm trajectories, locating the arm in the visual field, and
learning the map between gaze direction and correct pointing
configuration of the arm. All learning was self-supervised solely
by visual feedback."
Would that be what you mean? In that case I propose that we contact
on May 26, 2004 and see if we agree who has to come up with a
bottle. In case we disagree - remote chance, but who knows - we may
appoint a judge. Who would be able and willing to objectively judge
the matter... Perhaps we could persuade Brian O. Bush, or else
Richard Weait?
> (Mark - if you're reading this, there's noooo pressure. Nope, none at
> all.... ;>
:)
By the way, I would accept other achievements of similar
complexity. But centering the bet on vision, learning and
recognition could well be a good idea.
Best,
Steve
P.S: I have that paper in gzipped postscript, if you want it for
reference.
----------------------------------------------------------------------
# sbolt@xs4all.nl # Steven Bolt # popular science monthly KIJK #
----------------------------------------------------------------------
------------------------------------------------------------------------
eGroups.com home: http://www.egroups.com/group/alt-beam
http://www.egroups.com
- Simplifying group communications
Home