Alt-BEAM Archive
Message #02904
To: Bob Shannon bshannon@tiac.net
From: rdraycott@CPL.co.uk
Date: Tue, 4 May 1999 10:30:38 +0100
Subject: [alt-beam] Re: common body
I feel that I perhaps did not express myself very well
so I will try again after clarifying my position
Beam tec :- anything to do with the development of and evolution of
devices following the beam philosophy ! So yes it is
Now, the fundamenental aspect of a creatures evolution is only partially
its behaviour. Real world animals developed their anatomical differences as
a result of the environment they where in. You only have to look at nature
individual and group behaviours are not the only issue, how far do you
think evolution would have got if the amoeba had been the common body and
the only aspect of change was behaviour , indeed as in nature the
anatomical structure of the animal dictates its behaviour (e.g. peacock ,
courting display compared with that of a scorpion).
You also fail to interpret the test requirements , ( I develop software so
do understand the issues of testing I do it daily!)
you suggest 2 sets of tests one for behaviour and one to apply to test the
bot / environment but make no attempt to show any way of interpreting one
set of results against the other!!
wasteful...........
the real answer if you want a valid set of tests is to combine both and
leave the body issue open.
And this all starts with a requirements specification with this as a
grounding it is acting as the environmental conditions.
you have to specify your expectations AND the circumstances surrounding
those expectations,
as I said before the mars lander cr*ped out and needed re programming
because the spec it was designed on didn't include that requirement.( they
just got lucky that they could fix it !!!!!)
the spec should read for example (Simple example )
environment 4ft X 4 ft table 1 inch surrounding wall 500 watt light source
(obstacles to be specified , you think of something , very shallow pyramid
slope, 3 or for sided ? , you get the idea )
1) bot should keep moving ( not too long a delay between triggers)
2) bot should not lock into repeated behaviour (e.g. stand still or just
spin in the strongest light source)
3) bot should not require intervention during a minimum time.
4) bot should avoid collision's that could be detrimental to survival or
cause 2 or 3.
etc etc
with a standardised set of test requirements it ensures that full
development is retained ( builders don't centre just on one aspect of the
design) and that both issues, behaviour and environment handling are
tested.
Rob D
AKA "Powerbuilder"
rdraycott@CPL.co.uk wrote:
>
> someone wrote
> > The common body eliminates a lot of variables while permiting a
> >direct evaluation of different sets of behaviors. This would
> >greatly accelerate behavioral development and robot evolution.
>
> This is fundamentally flawed, enviromental requirements are supposed to
> define the evolutionary development of a creature/BOT
> with this suggestion you could be potentialy restricting the evolution of
> BEAM tec down a dead end since as in life the configuration of the body
> structure can prove to be a descisive evolutionary development (eg the
Bat
> compared to say a shrew)
No, not at all.
First, the discussion above is not talking about BEAM tech at all.
Secondly the environmental requirments for the robots being compared are
exactly the same, its a trial of different behavior sets being tested in
a known environment.
> yes it will speed up behavioral development BUT at the expense ot the
> architectual and constructional evolution.
Thats where the second set of trials enters the picture. Once you have
a standard way to testing behavior sets, its a simple matter to test a
modified or new body against the 'standard' design.
> As yhe text said the merit In a common body approach is in terms of
> judging comparitivly the actions of one against the other.
> I get the impresion that while such an aproach is a worthy reasearch
issue
> it certainly cuts down the fun of the design and construction.
Why?
You can still innovate new designs, but now you have a way to directly
test
to see how much improvement (if any) your modifications made?
------------------------------------------------------------------------
eGroup home: http://www.eGroups.com/group/alt-beam
http://www.eGroups.com
- Simplifying group communications
Home