Alt-BEAM Archive
Message #04855
To: beam@sgiblab.sgi.com, "Zozzles T. Freep" zozzles@lanl.gov
From: Bruce Robinson Bruce_Robinson@bc.sympatico.ca
Date: Thu, 24 Jun 1999 17:25:58 -0700
Subject: [alt-beam] Re: Wow! Observations! (was: Ah! Blessed Skepticism!)
"Zozzles T. Freep" wrote:
>
> Understood, Bruce, but he -is- showing that implex appears to exist -- at
> least for the bicore.
Oh, I never had a problem with implex (or internal feedback, or whatever
you want to call it) when the motor was driven directly from the Nv
outputs. I would have been very surprised indeed if it did NOT show up.
Most of the chips we use have rather modest power output capabilities,
so there is a very good chance of overloading the inverter when the
motor stalls. And as Mark pointed out, this will affect the NEXT Nv.
Even if the inverters were not overloaded, it's to be expected that
motor noise may have some effect.
>Wilf wrote:
>> ...even if a small resistor is introduced in series with the powersupply
>> to reduce supply voltage under load (so much for powersupply feedback).
>
> isn't he confirming what you are thinking in the first place?
Yes, he is. But I suspect there is one experimental flaw in this test
(why am I arguing against myself, anyway). My theoretical calculations
(after Ben set me on the right track) indicate that drawing down the
supply voltage will have no effect if it happens at the start of a
neuron's active cycle. The greatest theoretical effect would happen in
the middle of a cycle.
It sounds as though Wilf was slowing down the motor through the whole
cycle. So perhaps "general implex" DOES occur, under certain
circumstances. I still maintain that it will be overwhelmed by
mechanical effects.
> Ain't actual experimentation grand? My hat is off (yet
> again) to Wilf!
Mine, too. And experimentation is absolutely essential -- but it ought
to be guided by theory. Otherwise, our experiments will merely be
"tinkering ad infinitum", to quote Steve Bolt.
> However, Wilf acknowledges what I consider to be the second part of the
> problem (and probably the more important) -- does implex actually help
> the robot in any way, is it valuable to the robot?
Absolutely, and that's another reason I don't have much use for the
"drawing down the voltage theory" of implex feedback. If it DOES exist,
and if practice follows theory, it will only go into effect if a leg is
suddenly grabbed in the middle of it's cycle. I have yet to imagine a
general situation where this would be useful to a walker. It may indeed
be helpful for a specific walker configuration, in a specific
enviroment.
> ... once you've got those values, how to measure, how to measure? I
> assert that it would take continuous, objective observation of your
> Jurassic Park to even begin to address this problem -- five minutes now
> and then won't do it.
Observation, with a notebook, a plan, and a shorthand terminology to
describe the robot and it's motions. A camera is essential, possibly a
video camera. Practice observing. Read one of Tom Brown Jr.'s books (I
suggest "The Search" or "The Tracker") to get an idea of what this kind
of observation is all about.
And when I build a "successful" walker, I intend to leave it functional,
not tear it down for parts (take note, Steve). Why? Because the next
generation is supposed to be an improvement. And how better to observe
subtle differences than to do a direct comparison?
And on the less serious side of the discussion, since my RJP is acutally
an RFP (Robot Forest Park), comprising 14 acres of trees, brush,
hillside, mineshafts, swamps, and other hazards, I will very likely need
to employ some of Tom Brown's tracking skills to find the little
darlings.
Take care,
Bruce
------------------------------------------------------------------------
eGroups.com home: http://www.egroups.com/group/alt-beam
http://www.egroups.com
- Simplifying group communications
Home