Alt-BEAM Archive
Message #07203
To: beam@sgiblab.sgi.com
From: "Timothy Flytcher" flytch@hotmail.com
Date: Sun, 31 Oct 1999 01:44:31 PST
Subject: [alt-beam] Re: aquabot
Hi Jim, the letters about your aquabot make me think that maybe water isn't
the best medium... I mean algy and slime don't grow in many liquids??? maybe
a light oil??? paraffin? I don't think that flammable would really be a
problem if you seal it??? I have not built one so I don't have any real
answers just ideas :)
but I know that a lava lamps fluid is flammable (don't ask) it's just
sealed...
Timothy...
>Couple more things. If your light is bright enough to charge your bot, it's
>probably bright enough to grow a nice, healthy crop of algae. I have tried
>chlorine and sea salt.
______________________________________________________
7204 Sun, 31 Oct 1999 02:25:05 -0700 [alt-beam] Elementary, My Dear Watson... beam@sgiblab.sgi.com Jean auBois Ok, I get it that microprocessors are ok for the "horse and rider" paradigm
in BEAM but very generally (!) most BEAMers avoid them because they are
relatively complex -- just the number of connections much less the need to
know how to program and how to actually operate such a device. However,
there is one area that seems to be in the dark -- elementary digital logic
and state machines.
As can be seen by the use of the '138 for driving motors or H-bridges, a
little bit of digital logic goes a long way. Also, the question of
"memory" has resurfaced for which there have been basically two answers --
it is easy enough to use a latch/flip-flop/counter to do this; the other
answer has been more of a question as to how it could be done more BEAMishly.
The philosophical basis for BEAM has been described as making something
relatively simple and yet fairly chaotic and see what happens. If it turns
out not to be boring, then "taming the creature" appears to be the next
step. I believe that there is no fundamental argument against using this
philosophy to produce memory or more complex behaviors, but I think we are
missing the boat by ignoring what could be done with just a chip or two
more. Yes, one can make a fairly useful sort of "memory" using an Nu
neuron but it has some crippling limitations: the actual delay is
difficult to tune and it eventually turns "off" which might be undesirable
-- you might want to remember something "forever". At least it is easily
triggered and it can be "cleared" without much in the way of circuitry.
Oh -- by the way -- there is a kind of BEAM latch. If you take an Nv
neuron built with a '14 (i.e. it MUST have hysteresis) and add a pull-up
resistor whose value is equal to the pull-down resistor then the input to
the gate returns to Vcc/2. Presuming the edges that come into the
capacitor produce a voltage at the input of the gate either above the high
threshold or below the low threshold the gate will be "set" or
"reset". After that occurs, the voltage merely wanders back to Vcc/2 and
never crosses the other threshold so the gate maintains state. The only
problem is that if the input isn't at Vcc or ground then it has to be
floating (that is, having a well-define other-going edge just flips the
gate's state so you wouldn't have gained anything).
Purely linear sequences can be produced merely by using an unterminated
chain of Nv neurons -- plop a single pulse into the first neuron and it'll
bop along down the chain until it falls off the end. One can use varying
delays for each of the neurons if circumstances require it but one must
remember that this can be both tiresome to tune and subject to
drift. Producing conditionals -- if then else sorts of decisions -- would
probably be somewhat hard to do unless you were willing to give in and use
some digital logic for a digital function. However, if one did use a
demultiplexer (splitter) then the state of a memory element could shunt the
pulse into one of two (or many) subsequent neuron chains. As a result,
complex behavior would depend on what was remembered.
The reason why I've even started this diatribe is that the LEGO people have
an insanely great TV ad for their MindStorms product (go to
www.legomindstorms.com, scroll down to and click on the New TV Commercial
link, find Clean Up Time, download an appropriate copy and watch
it.) Watch how quickly the little bugger goes about its work -- the
BEAM-Ant does useful work and all but the RCX processor has it beat.
Nonetheless, there is no reason why a robot that is merely an evolutionary
step (leap?) from the BEAM-Ant couldn't do the same stuff as fast using
just a little more electronics.
Just some random thoughts whilst the change from daylight time to standard
time occurred,
JaB
Home