Controlling disturbances

[BRUCE_STEPHANIE_ABBO]

[From Bruce Abbott (2010.05.26.11:05 EDT)]

Bill Powers (2010.05.25.0855 MDT) –

Fred Nickols (2010.05.25.0731 PDT)

FN: Regarding your steering wheel example and Dag’s
point about

someone tugging at the wheel, I would imagine that
the controlled

variable isn’t the angle of the steering wheel but,
rather, the

unimpeded or free control of the movement of the
steering wheel.

BP: It’s simpler that that, I think. The higher-level
system controls

the direction in which the car is going (or its position
in a traffic

lane). It does so by varying the reference signal being
send to a

lower control system that keeps the wheel at whatever
angle is

currently set as a reference level. When someone tugs at
the wheel,

the first error is at the lower level: the wheel moves a
little,

creating an error relative to the reference signal being
received

from the car-steering system. That immediately produces
opposing

forces from the muscles of the arms, before the steering
system

senses any error in the position of the car. If the
disturbance is

large enough to cause the wheel to move at all (that
wouldn’t take

much force), the car will start to veer off course, and
very soon the

higher system will see an error in the direction or the
position of

the car in its lane. The higher system will change the
reference

position for the wheel-control system, just enough to
keep the car in

its lane even with the force still acting. A still-higher
system will

try to remove the disturbance, I would wager, by yelling
WHAT THE

HECK ARE YOU DOING??!!!. The person tugging on the wheel
has to be

careful when releasing the pressure, to allow the driver
to reduce

his opposing force slowly to zero. Letting go suddenly
would leave

the driver twisting the wheel to one side!

O.K. Bill, here’s a simplified
diagram of the car position-control system:

I’ve made the higher-level
system, controlling the position of the car relative to the lane, set the
reference for steering wheel angle, which in turn directly affects the car’s
position. In reality, steering wheel angle affects the direction and rate of
turn, which would be varied so as to change the car’s position (e.g., a little
left turn, then a little right turn, to move the car a bit to the left in its
lane). I wanted to avoid that added complexity to keep the diagram simple, saving
room for additional components to be added later.

Note the disturbance to the
steering wheel’s position that is being delivered via Bill’s arm muscles. What
I’d like to see added to the diagram is the higher-level system you described,
the one that, in your scenario, has the driver asking you what the heck you are
doing. I want to explore this model because it proposes something that seems a
bit different from the standard HPCT model: A system that detects the source of
a disturbance to another system and takes action to eliminate that source.

Bruce

[rsmarken]

[From Rick Marken (2010.05.27.0850)]

Bruce Abbott (2010.05.26.11:05 EDT)–

O.K. Bill, here’s a simplified
diagram of the car position-control system:

Very nice (or Van Nuys, as they say here in the borscht belt).

Note the disturbance to the
steering wheel’s position that is being delivered via Bill’s arm muscles. What
I’d like to see added to the diagram is the higher-level system you described,
the one that, in your scenario, has the driver asking you what the heck you are
doing. I want to explore this model because it proposes something that seems a
bit different from the standard HPCT model: A system that detects the source of
a disturbance to another system and takes action to eliminate that source.

That would be the Nazi control system, I believe;-) We all have one, of course, but it’s usually kept in tow by the “decency” control system, which, unfortunately, not everyone has. Both are part of the standard HPCT model but it’s not talked about much for obvious reasons.

Best

Rick

···

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[FredNickols]

[From Fred Nickols (2010.05.27.0954 PDT)]

I am glad Bruce A raised this issue because (at least at higher
levels of the hierarchy), I believe we often do exactly that – detect a
disturbance, identify its source, set a (establish a reference signal) for
eliminating that source, and act accordingly. I also believe we often act
to control perceptions of our behavior and act in ways that align our notion of
how we ought to be behaving with the way we see ourselves as behaving.
So-called “external feedback” can of course pose a disturbance to
that controlled variable.

Regards,

Fred Nickols

Managing Partner

Distance Consulting LLC

1558 Coshocton Ave - Suite 303

Mount Vernon, OH 43050-5416

www.nickols.us | fred@nickols.us

“Assistance at a Distance”

···

From: Control Systems
Group Network (CSGnet) [mailto:CSGNET@LISTSERV.ILLINOIS.EDU] On Behalf Of Richard
Marken
Sent: Thursday, May 27, 2010 8:48 AM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Controlling disturbances

[From Rick Marken (2010.05.27.0850)]

Bruce Abbott (2010.05.26.11:05 EDT)–

O.K. Bill, here’s a simplified diagram of the car
position-control system:

Very nice (or Van Nuys, as they say here in the borscht
belt).

Note the disturbance to the steering wheel’s
position that is being delivered via Bill’s arm muscles. What I’d like to see
added to the diagram is the higher-level system you described, the one that, in
your scenario, has the driver asking you what the heck you are doing. I want to
explore this model because it proposes something that seems a bit different
from the standard HPCT model: A system that detects the source of a disturbance
to another system and takes action to eliminate that source.

That would be the Nazi control system, I believe;-) We all
have one, of course, but it’s usually kept in tow by the “decency”
control system, which, unfortunately, not everyone has. Both are part of the
standard HPCT model but it’s not talked about much for obvious reasons.

Best

Rick

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Bill_Powers1]

[From Bill Powers (2010.05.31.0650 MDT)]

Bruce Abbott (2010.05.26.11:05 EDT)]

BA: O.K. Bill, here’s a
simplified diagram of the car position-control system:

Note the disturbance to the steering wheel’s position that is being
delivered via Bill’s arm muscles. What I’d like to see added to the
diagram is the higher-level system you described, the one that, in your
scenario, has the driver asking you what the heck you are doing. I want
to explore this model because it proposes something that seems a bit
different from the standard HPCT model: A system that detects the source
of a disturbance to another system and takes action to eliminate that
source.

BP: I see that I’ve failed to get my point across about sensing
disturbances, and have left a false impression that in PCT, disturbances
are NEVER sensed. I’ve been trying to counteract a model (mainly Ross
Ashby’s, but existing in many forms since then including what is called
“modern control theory”) in which sensing disturbances is the
only basis of control and is used without depending on negative feedback
at all. I’ve therefore put a lot of effort into showing that control can
easily take place without sensing the cause of the perturbations of the
controlled quantity. That proposition, if proven, destroys the Ashby and
MCT models by showing that they can’t work if the disturbance is known
only through the occurrance of errors. A driver, for example, could not
keep a car on the road in a crosswind, because the crosswind itself can’t
be sensed directly, and even the “cues” it generates, such as
blowing scraps of paper, can’t give enough quantitative information to
allow successful steering.
Now ask yourself if this really forbids sensing the causes of
perturbations. Note, please, that I’m using the word perturbation to mean
the effect of a disturbance. I promise to try never to use
“disturbance” to mean the effect, such as a change in a
controlled variable. If a controlled variable suddenly departs from its
reference level, that is NOT a disturbance. It is an effect of a change
in a disturbing variable somewhere else – it’s a perturbation in the
controlled variable, a change caused by something other than the action
of the control system itself

When I do use the word “disturbance,” therefore, I am always
referring to some independent variable that, through intervening
properties of the environment, can cause the controlled variable to
change.

The perceptual signal at the top left of your diagram is shown next to
the corresponding reference signal, labeled “Desired car position
(reference)”. The perceptual signal, then, must represent
“Perceived car position”. Besides entering the local comparator
function, it will follow a branch going upward to higher
systems.

Also, you show “Bill’s muscles”, but not “Dag’s
perceptions of Bill’s hand pulling on the steering wheel.” Without
trying to answer all the “how” questions, I think we can agree
that Dag would have perceptions of Bill’s hand, perceptions that are
going upward to higher systems along with perceptions of the car’s
position, and that at a higher level he might well perceive a
relationship of the former to the latter. And it’s not unreasonable, I
think, to propose that he has ways of influencing Bill’s behavior by
means of saying words with intonations, including the first thing Dag,
with immense restraint, said as I recall it: “What are you
doing?” He didn’t yell at me to stop, he first tried to determine
why I was doing it, even while he wrestled against the perturbation in
car position. He may have been a little too preoccupied to notice that
when I both started and stopped doing it, I applied and released my
disturbing force gradually, making sure he had control all the way so the
car didn’t actually depart from a safe position to an important degree.
Novices: do not try this at home.

Also, please note that what I did was not particularly dangerous, since I
had hold of the wheel and was attending with great interest and no
death-wish to the position of the car in its lane (I was in it). What I
was doing was no more dangerous than what Dag was doing, which was
deciding not to steer the car into approaching traffic or stationary
obstacles, as we all decide not to do every second we are driving. Tom
Bourbon was the first of us to notice how astonishingly low the accident
rate is on heavily-traveled roads. It looks almost as if people are
actively avoiding collisions.

Yes, to shorten this story, we can learn to oppose perturbations by means
of affecting the cause of the perturbation, the disturbing variable. That
makes the disturbing variable into part of the environmental feedback
function – a dependent variable now. The real disturbance now becomes
whatever it is that is causing the disturbing variable to change.
However, we don’t do this by calculating how much we have to change the
disturbing variable in order to remove the perturbation. We watch the
perturbation all the time, and apply our own influences to the (former)
disturbing variable as required to keep the perturbation as small as
seems necessary. We will now find that our actions have become equal and
opposite to whatever forces are making the former disturbing variable
change, plus enough more to remove any static effects.

Good question.

Best,

Bill P.

[FredNickols]

[From Fred Nickols (2010.05.31.0913 PDT)]

Thanks to Bill P and Bruce A for this exchange. It plugs a
long and vaguely-sensed gap. The distinction between perturbation and
disturbance is especially helpful.

Fred Nickols

fred@nickols.us

···

From: Control Systems
Group Network (CSGnet) [mailto:CSGNET@LISTSERV.ILLINOIS.EDU] On Behalf Of Bill
Powers
Sent: Monday, May 31, 2010 6:49 AM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Controlling disturbances

[From Bill Powers (2010.05.31.0650 MDT)]

Bruce Abbott (2010.05.26.11:05 EDT)]

BA: O.K. Bill, here’s a simplified diagram of the car
position-control system:

Note the disturbance to the steering wheel’s position that is being delivered
via Bill’s arm muscles. What I’d like to see added to the diagram is the
higher-level system you described, the one that, in your scenario, has the
driver asking you what the heck you are doing. I want to explore this model
because it proposes something that seems a bit different from the standard HPCT
model: A system that detects the source of a disturbance to another system and
takes action to eliminate that source.

BP: I see that I’ve failed to get my point across about sensing disturbances,
and have left a false impression that in PCT, disturbances are NEVER sensed.
I’ve been trying to counteract a model (mainly Ross Ashby’s, but existing in
many forms since then including what is called “modern control
theory”) in which sensing disturbances is the only basis of control and is
used without depending on negative feedback at all. I’ve therefore put a lot of
effort into showing that control can easily take place without sensing the
cause of the perturbations of the controlled quantity. That proposition, if
proven, destroys the Ashby and MCT models by showing that they can’t work if
the disturbance is known only through the occurrance of errors. A driver, for
example, could not keep a car on the road in a crosswind, because the crosswind
itself can’t be sensed directly, and even the “cues” it generates,
such as blowing scraps of paper, can’t give enough quantitative information to
allow successful steering.
Now ask yourself if this really forbids sensing the causes of perturbations.
Note, please, that I’m using the word perturbation to mean the effect of a
disturbance. I promise to try never to use “disturbance” to mean
the effect, such as a change in a controlled variable. If a controlled variable
suddenly departs from its reference level, that is NOT a disturbance. It is an
effect of a change in a disturbing variable somewhere else – it’s a
perturbation in the controlled variable, a change caused by something other
than the action of the control system itself

When I do use the word “disturbance,” therefore, I am always
referring to some independent variable that, through intervening properties of
the environment, can cause the controlled variable to change.

The perceptual signal at the top left of your diagram is shown next to the
corresponding reference signal, labeled “Desired car position
(reference)”. The perceptual signal, then, must represent “Perceived
car position”. Besides entering the local comparator function, it will
follow a branch going upward to higher systems.

Also, you show “Bill’s muscles”, but not “Dag’s perceptions of
Bill’s hand pulling on the steering wheel.” Without trying to answer all
the “how” questions, I think we can agree that Dag would have
perceptions of Bill’s hand, perceptions that are going upward to higher systems
along with perceptions of the car’s position, and that at a higher level he
might well perceive a relationship of the former to the latter. And it’s not
unreasonable, I think, to propose that he has ways of influencing Bill’s
behavior by means of saying words with intonations, including the first thing
Dag, with immense restraint, said as I recall it: “What are you
doing?” He didn’t yell at me to stop, he first tried to determine why I
was doing it, even while he wrestled against the perturbation in car position.
He may have been a little too preoccupied to notice that when I both started
and stopped doing it, I applied and released my disturbing force gradually,
making sure he had control all the way so the car didn’t actually depart from a
safe position to an important degree. Novices: do not try this at home.

Also, please note that what I did was not particularly dangerous, since I had
hold of the wheel and was attending with great interest and no death-wish to
the position of the car in its lane (I was in it). What I was doing was no more
dangerous than what Dag was doing, which was deciding not to steer the car into
approaching traffic or stationary obstacles, as we all decide not to do every
second we are driving. Tom Bourbon was the first of us to notice how
astonishingly low the accident rate is on heavily-traveled roads. It looks
almost as if people are actively avoiding collisions.

Yes, to shorten this story, we can learn to oppose perturbations by means of
affecting the cause of the perturbation, the disturbing variable. That makes
the disturbing variable into part of the environmental feedback function – a
dependent variable now. The real disturbance now becomes whatever it is that is
causing the disturbing variable to change. However, we don’t do this by
calculating how much we have to change the disturbing variable in order to
remove the perturbation. We watch the perturbation all the time, and apply our
own influences to the (former) disturbing variable as required to keep the
perturbation as small as seems necessary. We will now find that our actions
have become equal and opposite to whatever forces are making the former
disturbing variable change, plus enough more to remove any static effects.

Good question.

Best,

Bill P.