Alt-BEAM Archive
Message #02448
To: Steven Bolt sbolt@xs4all.nl
From: Bob Shannon bshannon@tiac.net
Date: Sat, 17 Apr 1999 21:33:32 -0700
Subject: [alt-beam] Re: Genome, CPU vs BEAM vs Hybrid, Copyrights,
Steven Bolt wrote:
>
> On Fri, 16 Apr 1999 JVernonM@aol.com wrote:
>
> > In a message dated 4/16/99 6:37:41 PM Eastern Daylight Time,
> > chiumanfu@home.com writes:
>
> > > For now I'm going to sidestep all the "other" topics cuz I
> > > have a lot of brainstorming to do. I have to think of
>
> What I'd like is:
>
> > > "a contest aimed at promoting task-oriented innovation,
>
> *With*
>
> > > a clear winner criterion - like task performed and fastest"
>
> > > How the hell do you do that without limiting the contest to
> > > just the upper echelon of BEAMers...and over the net too?
By exchanging sets of behaviors between a common development platform
and directly comparing them and selecting the most sucessful approaches!
> I can imagine more than one contest, each being defined by a
> certain easy to build `arena' and a task. Ideally there would be a
> range of tasks, starting simple - like building a photovore which
> has to reach the best lit part of a flat arena while dealing with
> solid cylindrical obstacles. At the next level, the obstacles may
> include less accommodating shapes and things like a sheet of
> transparent material with a fairly narrow passage on one side, and
> the way to the light may be literally an uphill struggle.
This is a very good start!
> At other levels there may be targets marked in certain ways (balls
> of conductive alu foil, for instance), which have to be gathered
> and moved to the best lit corner. Tasks of the latter kind may be
> performed by one robot, or by several working together; what counts
> is speed and gathering *all* targets.
>
> > The project should be done by the more advanced beamers.
>
> I should hope not. Imho the idea of BEAM is that everyone can
> participate in everything.
Actually the idea of the arena was not really BEAM at all!
Its an alternative to the current process of BEAM evolution, intended
to deliver much more highly evolved robots in less time, and for less
money spent.
After all, our time and money are limited assests, and we should make
the most of them.
> > As others advance, they could contribute as well. The contest
> > should be task oriented, but WITH a clear criteria for winners of
> > design, not winners who built them. Look, you have 5 advanced
> > beamers build a common body and brain combination.
>
> I don't believe in this common body/brain idea. The only things
> defined should be the arenas and tasks. Participants should
> innovate as they see fit, and not be obliged to use a fixed set of
> modules or components, or serve to glorify others. The only
> requirement is for each to build his/her own arena, measure
> performance and report if it's any good. The new and old parts of
> designs are so far generally obvious, and most builders state what
> they borrowed from predecessors. That seems good enough to me.
Without a common development platform you cannot easily compare
different
sets of behavior and problem solving strategies with all other variables
being equal.
This also gives us a force multiplication effect, every bit of code I
write
and share is useful to other developers without needed to adapt it to
their
specific hardware.
Not that this should preclude the development of new and improved common
development platforms through the same process of competition and
selection.
Simply developing a properly designed kit can provide this development
platform.
> > The "winning" designs are incorporated into the following years
> > bots with another agreed upon improvement.
>
> Agreed upon? Some designs will no doubt be immitated, but the idea
> is to get more innovation. Performing a task faster is sufficient
> indication of progress, better design elements will automatically
> survive the failures. No need to formally agree. Evolution may be
> lots of things, but bureaucratic it ain't...
This does not need to be bureaucratic at all!
The Photopopper has gone through several revisions. I see no reason why
the
common development platform would not also do the same, adopting the
more
sucessful modifications to last years models, as proven by competitons.
But being CPU and software based, this is not a BEAM thing at all, and
I'm sure quite a few BEAM beginners would choose not to learn how to
program and develop new behaviors.
But these people could still buy the kits, and use software that is
developed
by this process of competitive evolution.
You may not beleive in the common platform concept, but if you review
the
history of the personal computer you will see that this is an
inexcapable
evolutionary path in simple economic terms.
Its probably a lot easier to learn PBASIC or something than lean enough
to
design a new and more efficient solar engine. The common deveopment
platform
could be as easy to program as a STAMP.
No expensive programmers needed, no fancy software, its really not an
obstacle
anymore.
------------------------------------------------------------------------
eGroup home: http://www.eGroups.com/list/alt-beam
Free Web-based e-mail groups by eGroups.com
Home