Alt-BEAM Archive
Message #05507
To: beam@sgiblab.sgi.com
From: Richard Piotter richfile@rconnect.com
Date: Sat, 31 Jul 1999 21:50:29 -0500
Subject: [alt-beam] Re: BEAM and CPU's
What you describe is what I wanted to do with Challenger...
Unfortunately, the large frame is a bit expensive to motorize! I
intended to have a sort of "spine" that leads to each motor driver
input. There would be plugs on that "spine" that would accept boards.
each board would have a function (or functions). There would be an
inhibit signal that would lock the other boards off as long as it was
running (some boards however could simply not recieve the inhibit signal
so multiple boards can carry out different tasks). The theory was great,
but I could never have afforded all the motors. I wanted space so I woul
have room for upgrades. The tops of each board would have an inhibit
out, inhibit in, a pulse initiate, and a recycle (pulse output). Higher
controllers would only have to send pulses to each board and properly
inhibit others.
Unfortunately, I can't afford 9 high power/torque motors, nor the
batteries to run them. Nuts!!!
Quadrapod took another route. it recieves a constant input from a 6 Nv
net, and re-routes the signals via logic gates. it has 5 inputs and 7
gaits. Again, the motors stop me from building. I suppose I could try
with my BG Micro motors, but they are rather small for what I had
planned. Besides, I still have other robots to build. I'll have to build
a BEAM controller for my new Hexapod!
The pics are only a day old! I put them up yesterday! Scratch built 3
motor hexapod using nothing but clothes hanger wire and some cheap
hinges from the hardware store! Beats Lynxmotion and their $175 price
tag, or whatever the heck it is! Not bad!
Bruce Robinson wrote:
>
> Ian Bernstein wrote:
> >
> > ... It also seems that as soon as you want to add multiple sensors and
> > give the bot more gaits the circuit just grows dramatically and becomes
> > very complex.
>
> True ... but you might be able to simplify the complexity by looking at
> human motor skills.
>
> At the lowest level, we use the same motor neurons and feedback sensors
> (proprioceptors) to control all our physical skills. When we learn new
> skills, at first the brain tries to control all the muscles
> individually. That's why people are often awkward and clumsy when
> learning new tasks -- their brains get overloaded managing low-level
> actions. But when a task is performed often enough (approximately 50
> times or so), intermediate neurons are permanently programmed to
> co-ordinate muscle groups (Central Pattern Generators). Then the brain
> has only to send a few signals to the intermediate neurons, which in
> turn take care of co-ordinating very skilled tasks.
>
> Assuming we builders bypass the motor-learning process and just build
> the intermediate circuits, there's a couple of routes we could take.
>
> One BEAM equivalent would be to use a set of independent Nv networks to
> perform individual tasks; connect neurons from the different nets to
> individual motor inputs by using AND or NAND gates, and only operate one
> Nv net at a time. You could, for example, use one Nv net for a simple
> walking gait, a second for a running gait, and a third (on a complex
> 'bot) to walk backwards. This would be a very good method when you
> wanted to change the sequence that various motors operated in.
>
> A second BEAM equivalent would be to use completely different resistor
> values on the Nv neurons. Each set would be connected to a common
> terminal point via diodes. Setting a terminal point to ground would
> activate the set of resistors attached to it. Setting the terminal point
> to Vcc would inactivate the resistors. This would be a good method when
> you wanted to use the same sequence to operate motors, but vary the
> timing.
>
> Both these methods add to the complexity and parts count. The big plus
> is that they allow different functions to be individually fine-tuned.
>
> This idea can be carried to a higher level. Say I have separate Nv nets
> to walk forward, back up, turn quickly left, and run forward. Now I can
> create a higher level Nv sequence that in turn triggers the "back up"
> net, the "turn left" net, and the "run forward" net. So if my
> hypothetical 'bot senses a dangerous situation, it can activate the
> higher level loop that will cause it to back away, turn, and run in the
> direction it came from.
>
> Sounds good in theory -- will it work? I dunno. I'm going to find out,
> but probably not before Christmas. Wouldn't it be easier to program a
> CPU? Probably. Would making a CPU 'bot be as much fun? Nope, not for me
> .
>
> > I'm up to the challenge and just wanted to see what you guys thought
> > before I start. Also how many motors should I use, 4?
>
> Go for three -- two for the legs, one for the waist. Then we can compare
> results .
>
> Regards,
> Bruce
--
Richard Piotter
richfile@rconnect.com
The Richfiles Robotics & TI web page:
http://richfiles.calc.org
For the BEAM Robotics list:
BEAM Robotics Tek FAQ
http://people.ne.mediaone.net/bushbo/beam/FAQ.html
5508 Sat, 31 Jul 99 21:56:32 -0600 [alt-beam] Re: BEAM and CPU's Ian Bernstein Sounds good. After reading my message again I think maybe I' didn't
really get into my idea of buiding a basic master/slave bicore walker and
doing all the brain stuff with a cpu. Example: If you've ever seen a
spiders leg just after it was disconnected from the body it will keep
moving (bicore) but without the brain (cpu) it can't do anything. (I'm
not sure if any of that is correct but I'm just giving it as an example).
>Go for three -- two for the legs, one for the waist. Then we can compare
>results .
Just for the records. Mark T. has a 2 motor walker that can make a pivot
turn when it bumps an object so turning is possible.
Laterz
*-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-*
Ian Bernstein "aka - Synet" The Master Builder
E-Mail - Ian@beam-online.com
BEAM Online - http://www.beam-online.com
Quote - "Programming today is a race between software engineers
striving to build bigger and better idiot-proof programs, and the
Universe trying to produce bigger and better idiots. So far, the Universe
is winning."
- Rich Cook
Home