Alt-BEAM Archive

Message #00419



To: beam@corp.sgi.com
From: Terry Newton wtnewton@nc5.infi.net
Date: Mon, 15 Feb 1999 01:00:04
Subject: [alt-beam] BEAM, CPU's and the future


OK, you guys are doing it again...

CPU controlled robots, particularly those using tiny low-power
controllers like PICs, ARE often as simple and cheap as traditional
BEAM designs (besides being smarter) IF you have the programmer and
computer and software and the knowledge it takes to work with them.
If you don't, go analog, that's fun too! No need to argue about what
is better. Pointless unless the application is also specified. If
your goal is to pop to the light, no need for a cpu. If you want a
robot that can maneuver out of anything and remember how it did it,
I'd strongly advise learning how to use a microcontroller.

My (freeformed) PIC-based "popper" is entirely light powered, can pop
forward and reverse with one or both motors, and has a limited memory.
Provided nothing hangs up its tiny "wheels" it never gets stuck in a
corner. The maneuvering part of it could be done without a cpu but it
was convenient to use an 18 pin PIC and get sensor processing, memory
and future smarts upgrades for practically nothing.

My current design uses a PIC16C56 with an external eeprom - cost
about $6 for both. A PIC16LF84 has a built-in eeprom and runs down
to 2 volts, about $7 and save 1/2". The real cost of using cpu's
isn't in the processors and parts, they're cheap enough, it's in
aquiring the capability to work with them. Take me... my programmer
only does 5x parts (it's very old) so to use the new 2 volt PIC
I'm going to have to upgrade or buy a new programmer. $30-$60.
But I only need do this once and then I can do whatever I want to
with the new chip. Brain and capabilities permitting of course.

Having said all that, I think an analog (aka beam) reversible
photopopper would rock! forget about the memory, we already kinda
know what the feelers should do. I posted a rough design for one
to this list months ago but it was much more complicated than the
cpu-controlled design. Has anyone made a reversible popper-style
robot without a cpu?

BEAM changed my way of thinking when it comes to robots. My first
attempts were battery power hogs, complete with passive regulators
and relays wasting many hundreds of milliwatts. The current PICBOT
computer draws 1ma at 3 volts while popping, sleeps between each pop
dropping current down to a few microamps. It does need a bit of bright
light to come alive, seems the PIC powers up doing everything but
sleeping, but once the program gains control it conserves the charge
it has even overnight. Although the circuits aren't beam per say, the
specifications are inspired by BEAM doctrine. (besides the part about
not using a cpu... skipped that part:) This I believe will be the real
impact of BEAM technology, not so much the circuits themselves but
bringing up future engineers to think about efficiency, to think about
size, mechanics, survivability and the easiest way to get the job done.
While pure BEAM circuits might have limitations, the broader concepts
are just as applicable to conventional designs. BEAM teaches us to
imitate biology with electronics, make it simple and pleasing to the
eye (won't try to spell the A word right now...) and to pay attention
to mechanical design. How we do that is up to each of us, builders of
robots. We can be the future of robots. If it moves away from traditional
BEAM, that's just proof that the concepts are more than a few circuits.
If it stays with tradition, that's because those simple circuits are
pretty cool and do a lot for what little is there. I suspect the future
includes both. It is not either/or, it's whatever you want to do.

Whether or not I actually use "pure" circuits, I credit BEAM and
the visions of Mark Tilden and other BEAM experimenters for helping
me to see the light. Efficiency! Carry on...

Terry Newton


------------------------------------------------------------------------
eGroup home: http://www.eGroups.com/list/alt-beam
Free Web-based e-mail groups by eGroups.com

Home