[spam] Re: More on Disturbances

CSGnet Archive CSG2006
1

[Martin Taylor 2006.12.11.09.55]
To Fred Nickols apparently Mon, 11 Dec 2006 08:10:10 -0500

My take on Fred's questions, the last one first:

Earlier, I used a sunlight example and
indicated that I had to squint (whether to continue driving safely or simply
to protect my eyes needn't detain us at this point). The sun, if I
understand Bruce Nevin's point, is the "source" of the disturbance. The
"disturbance" itself is the effect of the sun's rays on my retina. Do I
have that right?

My take is that the "disturbance" itself is the effect the sun's rays have, not on the retina, but on whatever perception(s) you are controlling. I doubt you are controlling a perception of the sun's rays hitting your retina. You are probably controlling perceptions of things relating to your driving, some of which are subject to disturbances related to the effect of the sun on your retina. Disturbances are disturbances to perceptions, usually occasioned by physical effects on the snesory organs.

It seems to me there are two classes of disturbance: (1) those that we are
consciously aware of and attend to as such and (2) those that are simply
compensated for in our behavior with no thought to disturbance as such.

I would say that we are never conscious of disturbances as such. What we are conscious of is perceptions. If we don't compensate for independently caused effects on our perceptions, those perceptions change. Some of those changed perceptions may be ones for which we have reference values.

If we do have a reference value for a perception, and some external effect influences it to change, then either we have reorganized in such a way that our actions compensate against the disturbance (your class 2), or we have not reorganized effectively and we must imagine a way to return the perception to its reference value (your class 1).

Even though a control system whose perception is being disturbed cannot see the source of the disturbance, other systems may be able to see both the source and the effect of the source on the disturbed system. It is those other systems that come into play in ...

The first kind of disturbance, the kind we view as such and to which we give
deliberate thought as to the means of compensation, is perhaps represented
by a businessman's response to the successful launch of a new product by a
competitor. The businessman is controlling for factors such as sales and
profitability and things were going along just fine. But, the competitor's
new product takes away sales and profits and requires a response of some
kind so as to restore sales and profits.

A new and different issue is suggested by...

The second kind... Another example from my Navy days: I spent 14 years on sea
duty, some of it in rough weather. As I walked down the deck, whether going
forward or aft, the rolling and pitching of the ship I had to adjust my gait
so as to maintain my balance and forward movement. ("Sea legs" is the
common term and I developed the rolling gait of many a sailor.) It is only
with hindsight and knowledge of PCT that I conceptualize a disturbance in
that experience.

But when you came to land, did you not at first experience the land as swaying in the same way the ship did when you first put to sea? I certainly do after a time at sea. Where is the disturbance there? It's actually generated by your own (now unnecessary) compensatory actions, isn't it?

I think this is an interesting question in itself, but one that may belong in a different thread.

Martin

2

[From Bill Powers (2006.12.11.1040 <ST)]

Martin Taylor 2006.12.11.09.55 --

My take is that the "disturbance" itself is the effect the sun's rays have, not on the retina, but on whatever perception(s) you are controlling. I doubt you are controlling a perception of the sun's rays hitting your retina. You are probably controlling perceptions of things relating to your driving, some of which are subject to disturbances related to the effect of the sun on your retina. Disturbances are disturbances to perceptions, usually occasioned by physical effects on the snesory organs.

All right, I see you have chosen to go around again on this subject of disturbance as cause and disturbance as effect.

The ambient solar radiation falling on the eye is a disturbance having a certain magnitude measurable in watts per square meter. That is, it is a variable that has an independent effect on a controlled variable. It's a variable that is not affected by the action of the controller. The controlled variable is (variously) the contrast of the perceived image of a scene, or the detail perceived in the image. The effect of the disturbance is to reduce perceive contrast or detail below the desired level. We can, of course, refer to this change in perception as a disturbance, too, if we want to make sure that nobody can tell whether we're talking about the cause or its effect. Example:

No action you produce can alter the disturbance in any way. On the other hand, by taking certain actions such as shading your eyes, you can counteract the disturbance. So you can change the disturbance, even though the disturbance remains the same as it was before. In fact, if you control the controlled variable very tightly, applying even a fairly large disturbance will not produce a significant disturbance of the controlled variable. Given a large disturbance, in other words, there could be almost no disturbance.

To reply to Fred's comment about two classes of disturbance, I will maintain that there is only one class of disturbance: a variable having a state which can contribute to the state of a controlled variable. We never change the disturbance by our actions. Instead, we act to counteract its effect on the controlled variable. A crosswind, for example is a disturbance relative to a car's controlled direction of moment. We would probably measure the magnitude of the disturbance in kilograms of force applied to the car at right angles to its direction of movement. Turning the steering wheel to keep the car going straight does not affect the force of the wind on the car; the steering wheel simply adds a second force opposed to the first one. Relaxing the steering effort to let the crosswind blow the car off the road and onto an exit ramp does not affect the crosswind, either

If you somehow become able to affect the disturbance, it ceases to be a disturbance and becomes part of the feedback function through which your actions affect your perceptions. We would then change the definition of the disturbance so it once more becomes independent of your actions. For example if you held up an umbrella to shade your eyes, the disturbance would have to be redefined as the solar radiation arriving on the sunward side of the umbrella. Now you can alter the contrast of your perceived image by manipulating the umbrella, while the disturbance remains unchanged by your action. Using the umbrella, squinting, and holding a hand near your eyes are all methods used to control perceived contrast. They do not affect the disturbance at all.

All the confusion here is verbal.

Best,

Bill P.

3

[From Fred Nickols (2006.12.11.1603 EST)]

Bill Powers (2006.12.11.1040 <ST)]

We never change the disturbance by our actions. Instead, we
act to counteract its effect on the controlled variable.

That's what I thought.

A crosswind,
for example is a disturbance relative to a car's controlled direction
of moment. We would probably measure the magnitude of the disturbance
in kilograms of force applied to the car at right angles to its
direction of movement. Turning the steering wheel to keep the car
going straight does not affect the force of the wind on the car; the
steering wheel simply adds a second force opposed to the first one.
Relaxing the steering effort to let the crosswind blow the car off
the road and onto an exit ramp does not affect the crosswind, either

If you somehow become able to affect the disturbance, it ceases to be
a disturbance and becomes part of the feedback function through which
your actions affect your perceptions. We would then change the
definition of the disturbance so it once more becomes independent of
your actions. For example if you held up an umbrella to shade your
eyes, the disturbance would have to be redefined as the solar
radiation arriving on the sunward side of the umbrella. Now you can
alter the contrast of your perceived image by manipulating the
umbrella, while the disturbance remains unchanged by your action.
Using the umbrella, squinting, and holding a hand near your eyes are
all methods used to control perceived contrast. They do not affect
the disturbance at all.

All the confusion here is verbal.

I'll have to give all this some more thought, especially in relation to
actions consciously intended to counteract a disturbance and actions that do
so without conscious or deliberate thought behind them.

Regards,

Fred Nickols

4

[From Bill Powers (20067.12.11. 1425 MST)]

Fred Nickols (2006.12.11.1603 EST --

I'll have to give all this some more thought, especially in relation to
actions consciously intended to counteract a disturbance and actions that do
so without conscious or deliberate thought behind them.

The distinction here is between controlling something consciously and controlling the same thing unconsciously. There are many things we control either way. Conscious control seems especially necessary when we can't just use an automatic way of controlling, so some reorganization is needed.

But we don't control disturbances either consciously or unconsciously. We control perceptions of input quantities, not disturbances that tend to change those input quantities.

Perhaps what is confusing matters here is that we sometimes see an event about to happen that looks to us (at the program level) as if it may disturb something else that we're controlling. So we try to reason out an action that will prevent the effect of the disturbance from happening. If you're standing on the deck of a ship and see a huge wave approaching, you may deduce that when the wave gets to the ship an extra-large roll may occur and that you had better be ready if you don't want to lose your footing. So you grab hold of something and just before the wave arrives you lean opposite to the way you predict the deck is going to tilt. As a result, the wave has less effect on you than you thought it would. If, that is, you guessed correctly how much to lean, which way to lean, and when to lean.

Of course none of your actions had any effect on the disturbance. The wave kept right on coming, and when it reached the ship, the ship was raised and tilted just as if you had done nothing. Your actions altered the physical situation (you attached yourself to a railing or a stanchion) and you exerted forces that were roughly synchronized with the ship's roll. Those actions changed your behavior from what it was before, and as a result the disturbance was met by opposing forces. As an experienced sail,or, you might do all those things either consciously or unconsciously, meaning automatically the way you always do it, or conciously while you try to get better control.

Generally we oppose the effects of disturbance by producing our own effects on the controlled variable. Sometimes we use anticipation to bring the opposing action forward in time a little, but that approach has limited applicability because it depends on nothing unexpected happening. If you make the opposing move but the disturbance increases or decreases or changes direction or timing at the last second, your action may make the final result worse instead of better. On the football field it's called being "faked out of your shoes."

Here's a test of whether or not you are using the term "disturbance" in the PCT-correct way. A man walks up to the Great Pyramid of Giza, which weighs six gazillion tons, and leans against it with a total force of 20 pounds. How big is the disturbance of the pyramid?

Answer ______________

Best,

Bill P.

.

5

[From Fred Nickols (2006.12.11.1732 EST}] --

Here's a test of whether or not you are using the term "disturbance"
in the PCT-correct way. A man walks up to the Great Pyramid of Giza,
which weighs six gazillion tons, and leans against it with a total
force of 20 pounds. How big is the disturbance of the pyramid?

Answer ______________

The "disturbance of the pyramid" is zero because the pyramid is not trying to control any perceived variables.

The "disturbance" in relation to the man can't be determined because we don't know what variables he is trying to control.

Am I missing something here?

Regards,

Fred Nickols
nickols@att.net

6

[From Rick Marken (2006.12.11.1730)]

Fred Nickols (2006.12.11.1732 EST}] --

Bill Powers:
Here's a test of whether or not you are using the term "disturbance"
in the PCT-correct way. A man walks up to the Great Pyramid of Giza,
which weighs six gazillion tons, and leans against it with a total
force of 20 pounds. How big is the disturbance of the pyramid?

Answer ______________

The "disturbance of the pyramid" is zero because the pyramid is not trying to control any perceived variables.

The "disturbance" in relation to the man can't be determined because we don't know what variables he is trying to control.

Am I missing something here?

My guess is that the disturbance of the pyramid is the force that is "equal and opposite" to the force applied by the leaning man. I think the controlled variable would be the perception of leaning, which is mainly the perceived pressure on the shoulder. So for me:

Answer ___20 pounds_____

I hope I got this right because I'll be teaching a seminar on PCT next quarter at UCLA. If I didn't get it right I just won't bring up the Great Pyramid;-)

Best

Rick

···

---
Richard S. Marken Consulting
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

7

[From Bill Powers (2006.12.13.1625 MST)]

Rick Marken (2006.12.11.1730) –

My guess is that the disturbance
of the pyramid is the force that is “equal and opposite” to the
force applied by the leaning man. I think the controlled variable would
be the perception of leaning, which is mainly the perceived pressure on
the shoulder. So for me:

Answer 20
pounds__

By your definition, that should have been -20 pounds (equal and
opposite) if the man is pressing with a force of 20 pounds on the
pyramid. But even if you use the negative sign, your answer is not
correct.

Who said there is any controlled variable in this example? Are you (and
Fred) saying that the only variables that can be disturbed by another
variable are controlled variables?

Best,

Bill P.

8

[From Rick Marken (2006.12.13.1630)]

Bill Powers (2006.12.13.1625 MST)

Rick Marken (2006.12.11.1730) --

My guess is that the disturbance of the pyramid is the force that is "equal and opposite" to the force applied by the leaning man. I think the controlled variable would be the perception of leaning, which is mainly the perceived pressure on the shoulder. So for me:

Answer ___20 pounds_____

By your definition, that should have been -20 pounds (equal and opposite) if the man is pressing with a force of 20 pounds on the pyramid. But even if you use the negative sign, your answer is not correct.

Oh, rats! Can I have another guess? How about zero?

So, what is the answer?

Who said there is any controlled variable in this example? Are you (and Fred) saying that the only variables that can be disturbed by another variable are controlled variables?

Well, I must admit that I was thinking that the term "disturbance" referred to variables that influence a controlled variable. But, of course, that can't be the case because we talk about applying "disturbances" when doing the test for the controlled variable and, of course, those "disturbances" may be fully effective, indicating that the variable disturbed is not a controlled variable. So, I'm so sorry for ever think that. I'm going to go and wash out my mind with physics;-)

Best

Rick

9

[From Fred Nickols (2006.12.13.0920 EST)]

Hmm. I thought something was a disturbance (D) only if it is affecting a
variable (CV) that a control system is trying to control. You seem to be
saying that any variable (X) that affects any other variable (Y or Z) can be
considered a disturbance whether or not anyone is trying to control Y or Z.

In my own thinking, I have been concerned with consciously compensating for
disturbances of which I (or some other control system) are aware and that
are affecting a controlled variable. I guess I'll have to go back and
rethink all this.

Regards,

Fred Nickols
nickols@att.net

···

[From Bill Powers (2006.12.1615 MST)]

Back home after visit to daughter and grandsons in Durango.

Fred Nickols (2006.12.11.1732 EST}--

>The "disturbance of the pyramid" is zero because the pyramid is not
>trying to control any perceived variables.

A disturbance is a variable that contributes independently of the
action of any control system to the state of another variable. The
magnitude of the disturbance is the magnitude of the variable that is
doing the disturbing. Want to try again?

Hint: You're reading too much into the definition of a disturbance.

Best,

Bill P.

10

[Martin Taylor 2006.12.13.09.52]

[From Fred Nickols (2006.12.13.0920 EST)]

Hmm. I thought something was a disturbance (D) only if it is affecting a
variable (CV) that a control system is trying to control.

That has been the PCT-related concept, yes. But as the prior discussion brought out, it's not always clearcut.

  You seem to be
saying that any variable (X) that affects any other variable (Y or Z) can be
considered a disturbance whether or not anyone is trying to control Y or Z.

That would be the everyday concept of a disturbance, wouldn't it? If I understand Bill correctly, and I'm no longer sure I do, any variable may be influenced by an unknown set of influences, among which is included the output of a control system. Except for that single exception, the output of the control system, all the other influences are disturbances.

But that is a correct understanding, it raises another issue: viewpoint. Some variables are subject to influences from the outputs of many independent control systems. It is only from the viewpoint of a specific control system that one can say that there is a "single exception", which is its own output. The influences from othr control systems are disturbances to its controlled perception. From the observer's viewpoint, Bill's statement:

A disturbance is a variable that contributes independently of the
action of any control system to the state of another variable

would mean that the actions of one control system would not be considered disturbances to the perceptual signal of another.

This conundrum leads me back to the original concept, that a "disturbance" i the everyday sense means an influence that causes a variable to take on a different value than it would have in the absence of that influence. In the PCT sense, it's the totality of influences on a controlled perception other than the influence of the control system's own output.

In my own thinking, I have been concerned with consciously compensating for
disturbances of which I (or some other control system) are aware and that
are affecting a controlled variable.

"I" is not a simple control element. "I" is a complex of myriads of elementary control units (ECUs). The perceptual variable in an ECU is a scalar variable that represents the state of an observable that is influenced by disturbance and by the output of the ECU. The complex "I" control system may well contain an ECU "A" that observes the disturbance source of ECU "B" and that acts on the linkage between the disturbance source and the observable of ECU "A". "I", the totality of interlinked ECUs, can perceive lots of things at the same time. Some ECU(s) in "I" may well imagine the effects of their actions on their own perceptions of the linkages between the disturbace source and the observed variable of ECU "A". Those imaginings, together, might well constitute "consciously compensating for disturbances of which I (or some other control system) are aware and that are affecting a controlled variable."

Bill says:>

> Hint: You're reading too much into the definition of a disturbance.

but at the same time would like definitions to be precise so that we don't get into verbal confusions. I think that it is better that we do define it, but do so in terms of the canonical ECU loop diagram. In the loop diagram there is a simple loop perceptual singal -> error signal -> output signal -> input signal. Additionally there are two combining points: an internal one where the reference input is combined with the perceptual signal to form the error signal, and an external one where the influences of the output signal and another signal combine to generate the input signal. That "other" signal is the disturbance signal, by definition.

When you get into complexes of ECUs, such as "I", can you really talk of "a" disturbance? Each ECU's perceptual signal is potentially subject to influences other than those of its own output signal. And none of them can be "aware" of their own disturbances.

Martin

11

[From Bill Powers (2006.12.13.0947 MST)]

Aha! Martin joins the group being detained after school.

Martin Taylor 2006.12.13.09.52 –

That would be the everyday
concept of a disturbance, wouldn’t it? If I understand Bill correctly,
and I’m no longer sure I do, any variable may be influenced by an unknown
set of influences, among which is included the output of a control
system. Except for that single exception, the output of the control
system, all the other influences are disturbances.

In the ordinary language of physics, one variable that causes another to
change is a disturbance relative to that other variable. A change in
evaporation rate is a disturbance that tends to change the equilibrium
temperature of a lake. The gravitational field of Jupiter is a
disturbance that changes the orbit of Mars (more than the other planets’
orbits). The force (20 pounds) with which a man leans on a pyramid is a
disturbance that tends to deform the pyramid in the direction away from
the man.

Disturbance
Effect of disturbance
Evaporation rate
change
Change in equilibrium temperature
Gravitational field of
Jupiter Distortion of
orbit of Mars
Force applied to pyramid (by anything) Deformation (small) of shape of
pyramid
Independent variables (any number) Dependent
variable
Even physicists can be sloppy and use the word “disturbance”
when they mean “effect of a disturbance.” That is, they could
speak of a disturbance of the equilibrium temperature, a disturbance of
the orbit of Mars, or a disturbance of the shape of the pyramid. This way
of using the word encompasses both an unmentioned cause of the change in
the variable of interest, sort of off-stage, and the effect that is the
focus of attention. It squashes cause and effect together. This causes no
less confusion in physics than it does in PCT. The same thing, of course,
happens when we talk about behavior: is “setting a clock” an
action or a perception? Can muscles produce an output called
“setting a clock” or “orienting to a sound” or
“scratching an itch”?
When a control system is involved, its output variable, qo, is a
disturbance that affects the controlled variable along with effects of
many other disturbances. The effect of the control system’s disturbance
on the controlled variable, however, ends up automatically being equal to
and opposite to the sum of the effects of all the other disturbances.
When trying to make this all very clear (I hope), I usually use the term
“disturbing variable” rather than just saying
“disturbance,” because I know how other people use the
unadorned term. I give the disturbing variable a symbol, D, distinct from
the symbol that indicates the variable being affected by the disturbing
variable (qi). This was supposed to convey the idea that D is measured in
one location in the environment, while qi is measured in a different
physical location (usually). It is thus possible for D to change while qi
changes the same way, the opposite way, or not at all, depending on what
the disturbing variable altered by the control system (qo), and any other
disturbing variables, are doing at the same time.
As Martin indicates, in PCT we usually reserve the word
“disturbance” for a variable that affects the input quantity of
a control system independently of the system’s output quantity, and the
output quantity is called the system’s action, or just qo. Those terms
and distinctions are only conveniences, however. There is nothing special
about causation in a control system, or about the roles of dependent and
independent variables. We are not trying to develop a new jargon
independent of physics or other sciences. It’s just that when we give all
the variables the same kind of label – D1, D2, …Dn, where one of the
Ds is the output of the control system – it becomes difficult to
communicate that only one of the Ds is the action of the control system,
while all the others are something else. Calling the D that is produced
by the control system “qo” is meant as a mnemonic aid. The
“q” says this disturbance is a physical quantity
measured at a particular place in the environment, and the subscript
“o” says it is associated with the output of the system.
It is a variable of type q, subtype o for output. The physical
quantity associated with the input to the system is therefore “q sub
i”. It is also a variable of type q, but of subtype “i”
for input. In all this we are assuming the correctness of the physical
model of the environment.

In short, all we need to know about a disturbance is that it is a
physical variable linked by physical laws (or other laws, like social
laws) to another variable, so the disturbance is an independent variable
and the affected variable is a dependent variable. If we want to use this
term in connection with a control system, then we have to specify the
details: which effect of the disturbance is associated with the
controlled input quantity, and which other effect is due to the control
system’s own output quantity, the disturbance we give this special name,
qo, to single it out.

In my example of the man leaning on the pyramid, I said nothing about any
control system, either in the man or in the pyramid. The man was applying
a disturbance of 20 pounds to the pyramid. We might guess that the effect
of this force on the 20-gazillion-ton pyramid was quite small because the
pyramid is very heavy and made of rock. However, as Rick pointed out, we
could measure one effect of this disturbance very accurately, given
Newton’s Laws. The pyramid would be caused to push back with a force of
-20 pounds. But that is not a disturbance: it is an effect of the
20-pound disturbance. The reaction force is a dependent
variable.

So: in PCT we use the term “disturbance” for a cause, an
independent variable, and not for an effect or a dependent variable. We
use it in its ordinary physical sense, not in any special sense peculiar
to control systems. We give some disturbances special names like qo for
ease of communication but not because they have special
properties.

The answer in the blank part of my question is 20. Does everyone now
agree with that?

Best,

Bill

12

[From Rick Marken (2006.12.13.0945)]

Fred Nickols (2006.12.13.0920 EST) to Bill Powers

Hmm. I thought something was a disturbance (D) only if it is affecting a
variable (CV) that a control system is trying to control. You seem to be
saying that any variable (X) that affects any other variable (Y or Z) can be
considered a disturbance whether or not anyone is trying to control Y or Z.

For the reasons I mentioned in my earlier post -- specifically the fact, in the TCV, we apply "disturbances" to hypothetical controlled variables, which may, therefore, turn out to _not_ be controlled variables-- I think the term "disturbance" must refer to any variable that contributes to variation in another. So even the output of a control system is a "disturbance" if it influences the value of another variable; it's just a disturbance that is caused by the system itself.

In my own thinking, I have been concerned with consciously compensating for
disturbances of which I (or some other control system) are aware and that
are affecting a controlled variable. I guess I'll have to go back and
rethink all this.

I think this may be the real crux of the problem because it turns on the distinction between disturbance as cause of variation in a controlled variable (the PCT meaning of disturbance) and disturbance as an effect that shows up as a component of the variation in a controlled variable (the colloquial meaning of disturbance). This latter colloquial meaning of disturbance is described in PCT as variations in the controlled variable.

A control system perceives and can be conscious only of variations in the controlled variable; it doesn't perceive the disturbances that are the cause of variations in the controlled variable. So there is no way for a control system to consciously compensate for disturbances (where disturbances are the cause of variation in the controlled variable) because it doesn't perceive them. But the system can consciously compensate for disturbances in the (PCT incorrect) sense of effects on variations in the controlled variable. This is called "conscious control" because you are aware of the variations in the perceptual variable you are perceiving and controlling. But you are _not_ conscious of disturbances (PCT sense of cause of variations) to the controlled variable.

If control system B perceives and controls one of the variables that is a disturbance to the variable controlled by control system A, then system B can become conscious and control a disturbance (causes of variation) to the variable controlled by system A. But still what is being controlled by system B is the perceptual consequence of this disturbance variable as they show up as effects on system B's controlled variable.

A control system can never really consciously compensate for a disturbance (in the PCT sense of a variable that contributes to the variation in a controlled variable). You can see this by looking at the diagram of a control system. The only variable that a control system perceives (and, thus, can become conscious of) is qi, the controlled input variation. The disturbance, d, represents all causes of variation in qi other than the outputs of the system itself. The system doesn't perceive (so it can't become conscious of) d or o in the diagram.

Best

Rick

···

---
Richard S. Marken Consulting
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400