NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/bit/listserv/csgl/1390

Text File | 1992-11-06 | 5.5 KB | 131 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!news.gtech.com!noc.near.net!mars.caps.maine.edu!news.yale.edu!yale!yale.edu!jvnc.net!darwin.sura.net!paladin.american.edu!auvm!AERO.ORG!MARKEN
X-Delivery-Notice:  SMTP MAIL FROM does not correspond to sender.
Posted-Date: Thu, 05 Nov 92 10:59:23 PST
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.csg-l
Date:         Thu, 5 Nov 1992 10:59:23 PST
Sender:       "Control Systems Group Network (CSGnet)" <[e-mail]>
From:         [e-mail]
Subject:      PCT References, Re: Squirrels, airconditioners
Lines: 118

[From Rick Marken (921105.1100)]

Jeff Hunter (921104) --

I think it's great that you are trying to model control
systems. I would strongly suggest that you try to get
hold of Powers' series in Byte (June-Sept, 1979) and my
spreadsheet modelling paper; now available on Bill Silvert's
listsever but, EVEN BETTER, available in MIND READINGS from
CSG Publishing -- just send $18.00 to G. Williams,
460 Black Lick Rd., Gravel Switch, KY 40328 U.S.A.

In fact, I think this is a good time to remind all of you who
REALLY want to learn PCT that there is some VERY IMPORTANT
reference material available that should be disgested thoroughly
before making definitive statements about PCT and how it works.
Everyone should, of course, OWN (and re-read periodically) Powers'
classic Behavior: The control of Perception. I don't know the
ordering info on this but the publisher is Aldine/DeGruyter (which is
somewhere in NY). Every serious PCTer (especially would-be modellers)
should OWN a copy of every book published by CSG Publishing;
there are now four books avaialable from CSG Publishing; Powers'
Living Control Systems I and II, Robertson and Powers' Intro to
Psychology and Marken's Mind Readings.

Unfortunately, the Powers' Byte articles are not yet in any of these
books -- it is important to have copies of these, however.
---
Now, back to Jeff's post:

>        Now consider the squirrel trying to maintain a CEV
>"I am well fed". This is set by the squirrel's metabolism and diet.

A CEV is a variable. It sounds like you are describing the reference
for the variable "fed" -- the reference is "well".

>    The "gain" of the controlled system is (at most) -0.5, and
>indeed the error is often non-zero. However it keeps the environment
>within 8 units of zero 93% of the time.

You have ignored the dynamics here; this is, indeed, a low gain system
but it responds instantanously and completely to disturbance; it
also completes its action before the next disturbance can have any
effect. This unusual dynamic (compared to the way variables change
in the real world) is what is responsible for the relatively good
control exhibited by this low gain system; put such a system in
an environment of continuously changing variables (ie. the real world)
and add the proper dynamics to the system and you will see that a
system with this kind of gain don't control hardly much at all.
There are some good tips about computer modelling in the Byte articles.

>
> %wavers: loops=1 bias=0, incr=4
>
> %wavers: free max=  404 min= -124 sum=  208.36 err=51220.66
> %wavers: cont max=   28 min=  -30 sum=    3.20 err=   23.29
> %wavers: bias max=   61 min=  -13 sum=    6.67 err=  104.30
>

It's excellent to collect data like this. Bravo. But remember, these
results apply only for the discreet system (and environment) that
is instantiated in you program. Now try a system that is dynamically
realistic -- that STARTS responding to the current effect of the
disturbance on each iteration. Dynamics makes a BIG difference that
those of us reared on digital (as opposed to analog) computers often
tend to forget (like me -- I'm a digital baby, though I do have a dim
memory of analog machines).

>    The climate control ECS uses the input function:
>
>    delta heat = expected heat from concert * fudge factor -
>                  cooling capacity of air conditioners
>    delta temp = delta heat * thermal mass of the concert hall
>    predicted temp = delta temp + current temp
>
>    The comparator of the ECS is:

>        if (predicted temp >= max safe temp)
>            error =  predicted temp - max safe temp
>        else
>            error = 0
>
>    And the output function is:
>
>        if (error not zero)
>            turn on air conditioners
>


OK, this is a start. Now try to actually run the simulation (you'll
need to do quite a bit of work still to make it realistic -- espectially
putting in those pesky dynamics). But I think that you will get quite an
interesting surprise if you set up the model correctly. You will find
that your "predicted temp" function doesn't help much (compared to
just controlling without prediction) unless you are nice to the control
system and make actual variations in temperature (disturbances) a close
mimic of the predicted values. But convince yourself; feedforward
(which it what it seems you are trying to make a case for -- or
predictive control, or whatever you want to call it) doesn't help,
unless, of course, the prediction is correct. For some reason, the
idea of predictive control is quite seductive; the only way to convinve
oneself of it's uselessness is through ACCURATE modelling (I use emphasis
because with predictive control, it's easy to, well, cheat -- since you
can set things up to the benefit of the model).

Anyway, keep up the modelling.

Regards

Rick

     **************************************************************

Richard S. Marken                   USMail: 10459 Holman Ave
The Aerospace Corporation                   Los Angeles, CA 90024
E-mail: [e-mail]
[phone] (day)
[phone] (evening)