Classical Control Systems Lectures

[Abbott]

From Bruce Abbott (2014.12.18 1940 EST)

For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control” (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture ”examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link “41 videos” next to Brian Douglas’ name just below the video, on the left to return to the page on which all the videos are displayed.

Bruce A.

[wmansell]

Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero? Stability within PCT relies of a flexibly changing output to ensure that the INPUT remains as close as possible to the reference value for that input? My mood is stable not because I do nothing to change it but because it remains around a range I want it to be because of, and/or despite of, my behaviour?
Warren

···

On Thu, Dec 19, 2013 at 12:40 AM, Bruce Abbott bbabbott@frontier.com wrote:

From Bruce Abbott (2014.12.18 1940 EST)

For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control” (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture ”examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link “41 videos” next to Brian Douglas’ name just below the video, on the left to return to the page on which all the videos are displayed.

Bruce A.

–
Dr Warren Mansell

Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[Abbott]

From Bruce Abbott (2014.12.19 1020 EST)

Warren Mansell –

WM: Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero? Stability within PCT relies of a flexibly changing output to ensure that the INPUT remains as close as possible to the reference value for that input? My mood is stable not because I do nothing to change it but because it remains around a range I want it to be because of, and/or despite of, my behaviour?

Warren

I agree that Douglas’ way of defining stability is unfortunate. The problem you note arises partly because he is thinking of zero as the initial value (before the system is disturbed), but doesn’t mention that explicitly. A stable system will settle to (that same) zero after being disturbed. Also contributing to the problem is that Douglas is an control systems engineer and is using an engineer’s definitions of the various quantities. The output in a standard engineering diagram of a control system is what we would label the input. For example, the output of a voltage regulator is the regulated voltage. If the reference for the voltage is +5 Volts, then the regulated output should be as close as possible to + 5 Volts.

So when Douglas asserts that in a stable system, the output goes to zero, he is not saying that the actions go to zero, but that the variable is returning to its initial state.

So, what Douglas actually is asserting is that, in a stable system, after being disturbed the variable in question will settle back to its initial value.

If you continue watching the video, you will see an illustration in the form of a ball in a valley. Its “zero” is at the bottom center of the valley. If you disturb the ball (push it uphill and release) it will eventually come to rest back at zero (assuming there’s some energy loss due to friction of viscous damping). One could define the ball’s initial position as “3 inches from the left edge of the bowl.” In that case stability would be present if the ball eventually settled at 3 inches.

It’s important to note that Douglas does recognize that what is controlled in a control system is the system’s perception of the controlled variable. In the video, Robotic Car, Closed Loop Control Example, Douglas uses a simple tracked vehicle to illustrate the difference between open-loop and closed-loop control. The open-loop version can be aimed at a particular target location and will drive there so long as there are no disturbances. But when disturbances act to turn the car, the car does not reach the target. The closed-loop version has an accelerometer that senses the angular acceleration of the car in the horizontal plane. Beginning at 8:00 minutes into the video, Douglas diagrams the system. The output on his diagram is the actual angular acceleration of the car, but Douglas notes that the only thing the control system knows about the car’s angular acceleration is the sensed value provided by the accelerometer.

In engineering, the term “control” is used in a looser sense than we use the term in PCT. A motor “controller,” for example, may consist of a simple rheostat that varies the current flow to the motor. The rheostat is said to “control” the motor speed (in the sense of being an important factor determining that speed). In the absence of any disturbances to the motor’s speed, one could make up a table or formula giving the motor’s speed for any given position of the rheostat’s dial. Such a system may be described as employing “open-loop control.” It has no ability to counteract disturbances, so a load on the motor will slow it down below the rpm given in the table.

In the diagram of this system, there’s a reference (dial setting) the inputs to the “controller” and an output, the motor’s actual speed. To convert this system to closed-loop control, you read this output with a sensor and feed the sensor’s reading (after a change in sign to negative) into a comparator. The reference moves to the comparator and the output of the comparator, the error signal, now becomes the input to the “controller.”

So-called open-loop controllers are used by human beings, who may observe the actual output and adjust the reference (e.g., the rheostat’s dial setting) to compensate for any error between what the operator wanted and that the system is actually delivering. In so doing, the operator closes the loop. The term open-loop “controller” is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I think it makes sense in that context. In PCT, of course, we prefer to restrict the term “control” to systems that compare the actual state of affairs to the desired one and take action to reduce or eliminate any difference between them.

It’s important to understand these differences in terminology (and worth the effort to learn them) so that we are not led into believing that engineers are talking nonsense about control systems when in fact they are just using somewhat different terminology and diagrams than we use in PCT when saying the same thing.

Bruce

Bruce Abbott (2014.12.18 1940 EST)

BA: For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

BA: Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control” (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture ”examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link “41 videos” next to Brian Douglas’ name just below the video, on the left to return to the page on which all the videos are displayed.

[MartinT]

Warren,

I haven't looked at the links yet, but I suspect you are thinking of

“stability” differently from the way a control engineer would think
of it. To me, with my undergraduate engineering background,
“stability” means that the effects of a slight shift in any of the
signal variables will disappear over time, and all the variables in
the system will return to their previous state. Lack of stability
means that after such a slight change in a signal value, the system
will not by itself return to the original state, and in all
likelihood will continue to diverge further.

That's an oversimplification, because "state" is implicitly taken to

refer to static values, but it will do as a suggestion as to why you
might have not interpreted the lecture as it was intended. Stability
in a control loop refers to the condition of everything in the loop,
whereas in PCT, as you say, it refers to the ability of the loop to
return a specific value, the perception, to a predetermined level.
The two concepts are different.

But then again, I haven't seen the lecture.

Martin

···

On 2013/12/19 7:45 AM, Warren Mansell
wrote:

    Hi Bruce, I switched off after the first minute.

Surely, from a PCT perspective, stability cannot be defined as
the capacity to restore to an output of zero? Stability within
PCT relies of a flexibly changing output to ensure that the
INPUT remains as close as possible to the reference value for
that input? My mood is stable not because I do nothing to change
it but because it remains around a range I want it to be because
of, and/or despite of, my behaviour?
Warren

      On Thu, Dec 19, 2013 at 12:40 AM, Bruce

Abbott bbabbott@frontier.com
wrote:

                From

Bruce Abbott (2014.12.18 1940 EST)

                For

those of you who would like to learn more about the
classical approach to control systems, I highly
recommend a series of YouTube videos presented by
control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos
) . The lectures range
from basic concepts to sophisticated techniques used
in control system design, analysis, and tuning, so
whether you are a control systems neophyte or an
engineer who needs a bit of brushing up on these
techniques, you will find something worthwhile to
view.

                Douglas

is an engaging and talented teacher who is able to
communicate complex ideas clearly. A nice place to
start is with Douglas’ lecture, “introduction to
system stability and control” (http://www.youtube.com/watch?v=uqjKG32AkC4 ).
Another is the lecture ”examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM ).
Once you have watched a given video, simply click on
the link “41 videos” next to Brian Douglas’ name just
below the video, on the left to return to the page on
which all the videos are displayed.

Bruce A.

–
Dr Warren Mansell

      Reader in Psychology

      Cognitive Behavioural Therapist & Chartered Clinical

Psychologist

      School of Psychological Sciences

      Coupland I

      University of Manchester

      Oxford Road

      Manchester M13 9PL

      Email: warren.mansell@manchester.ac.uk

       

      Tel: +44 (0) 161 275 8589

       

      Website: [http://www.psych-sci.manchester.ac.uk/staff/131406](http://www.psych-sci.manchester.ac.uk/staff/131406)


        See [teamstrial.net](http://teamstrial.net) for further information

on our trial of CBT for Bipolar Disorders in NW England

        The highly acclaimed therapy manual on [              A Transdiagnostic Approach to CBT using

Method of Levels](http://www.amazon.co.uk/Transdiagnostic-Approach-Method-Levels-Therapy/dp/0415507642/ref=sr_1_1?ie=UTF8&qid=1351756948&sr=8-1) is available now.

        Check [www.pctweb.org](http://www.pctweb.org)
        for further information on Perceptual Control Theory

[wmansell]

Thanks Bruce! That starts to make a lot of sense, and also explains why things can get messy when I have talked to control engineers before, and also maybe explain why control theory applications to living systems got off to a difficult start with cybernetics? I will certainly watch further - sorry for my false start - as I do want to be able to communicate with real control system engineers - there are many here at the university I could be collaborating with. I am wondering though whether the human controller in the example you mention actually has a hierarchical relationship with the rheostat in controlling its reference value rather than making it open loop?
Warren

···

On Thu, Dec 19, 2013 at 3:23 PM, Bruce Abbott bbabbott@frontier.com wrote:

From Bruce Abbott (2014.12.19 1020 EST)

Warren Mansell –

WM: Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero? Stability within PCT relies of a flexibly changing output to ensure that the INPUT remains as close as possible to the reference value for that input? My mood is stable not because I do nothing to change it but because it remains around a range I want it to be because of, and/or despite of, my behaviour?

Warren

I agree that Douglas’ way of defining stability is unfortunate. The problem you note arises partly because he is thinking of zero as the initial value (before the system is disturbed), but doesn’t mention that explicitly. A stable system will settle to (that same) zero after being disturbed. Also contributing to the problem is that Douglas is an control systems engineer and is using an engineer’s definitions of the various quantities. The output in a standard engineering diagram of a control system is what we would label the input. For example, the output of a voltage regulator is the regulated voltage. If the reference for the voltage is +5 Volts, then the regulated output should be as close as possible to + 5 Volts.

So when Douglas asserts that in a stable system, the output goes to zero, he is not saying that the actions go to zero, but that the variable is returning to its initial state.

So, what Douglas actually is asserting is that, in a stable system, after being disturbed the variable in question will settle back to its initial value.

If you continue watching the video, you will see an illustration in the form of a ball in a valley. Its “zero” is at the bottom center of the valley. If you disturb the ball (push it uphill and release) it will eventually come to rest back at zero (assuming there’s some energy loss due to friction of viscous damping). One could define the ball’s initial position as “3 inches from the left edge of the bowl.” In that case stability would be present if the ball eventually settled at 3 inches.

It’s important to note that Douglas does recognize that what is controlled in a control system is the system’s perception of the controlled variable. In the video, Robotic Car, Closed Loop Control Example, Douglas uses a simple tracked vehicle to illustrate the difference between open-loop and closed-loop control. The open-loop version can be aimed at a particular target location and will drive there so long as there are no disturbances. But when disturbances act to turn the car, the car does not reach the target. The closed-loop version has an accelerometer that senses the angular acceleration of the car in the horizontal plane. Beginning at 8:00 minutes into the video, Douglas diagrams the system. The output on his diagram is the actual angular acceleration of the car, but Douglas notes that the only thing the control system knows about the car’s angular acceleration is the sensed value provided by the accelerometer.

In engineering, the term “control” is used in a looser sense than we use the term in PCT. A motor “controller,” for example, may consist of a simple rheostat that varies the current flow to the motor. The rheostat is said to “control” the motor speed (in the sense of being an important factor determining that speed). In the absence of any disturbances to the motor’s speed, one could make up a table or formula giving the motor’s speed for any given position of the rheostat’s dial. Such a system may be described as employing “open-loop control.” It has no ability to counteract disturbances, so a load on the motor will slow it down below the rpm given in the table.

In the diagram of this system, there’s a reference (dial setting) the inputs to the “controller” and an output, the motor’s actual speed. To convert this system to closed-loop control, you read this output with a sensor and feed the sensor’s reading (after a change in sign to negative) into a comparator. The reference moves to the comparator and the output of the comparator, the error signal, now becomes the input to the “controller.”

So-called open-loop controllers are used by human beings, who may observe the actual output and adjust the reference (e.g., the rheostat’s dial setting) to compensate for any error between what the operator wanted and that the system is actually delivering. In so doing, the operator closes the loop. The term open-loop “controller” is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I think it makes sense in that context. In PCT, of course, we prefer to restrict the term “control” to systems that compare the actual state of affairs to the desired one and take action to reduce or eliminate any difference between them.

It’s important to understand these differences in terminology (and worth the effort to learn them) so that we are not led into believing that engineers are talking nonsense about control systems when in fact they are just using somewhat different terminology and diagrams than we use in PCT when saying the same thing.

Bruce

Bruce Abbott (2014.12.18 1940 EST)

BA: For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

BA: Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control” (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture ”examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link “41 videos” next to Brian Douglas’ name just below the video, on the left to return to the page on which all the videos are displayed.

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist

School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[FredNickols]

[From Fred Nickols (2013.12.19.1243 EST)]

Consider the toys of my youth – the gunfire control systems on board US Navy destroyers.

The computer solves what is called the fire control problem which results in pointing a gun mount in a way that when a round is fired it will explode in the vicinity of the (moving) target.

One of the values the computer calculates is gun elevation order (i.e., the angle at which the barrel should be raised in the vertical plane). Two feedback loops come into play here.

First, there is the matter of positioning the gun elevation order module to the calculated value. The position being sent to the gun mount is by the gun elevation module is compared with the calculated value to determine if the proper gun elevation order signal is being sent.

Second, there is the matter of the gun mount’s actual position in relation to the ordered position. In this case, the receiver regulator in the gun mount compares the ordered position with the actual position.

Zero outputs indicate alignment of actual position with ordered or calculated position. BUT zero output is not what is being controlled (even though correspondence between actual and ordered is a goal). What IS being controlled is in the first case the ordered position being sent to the gun mount and in the second case the actual position of the gun mount in relation to the ordered position. In both cases what is being controlled is an input, not an output.

These gunfire control systems are designed and built by control system engineers.

Regards,

Fred Nickols, CPT

Managing Partner

Distance Consulting LLC

The Knowledge Workers’ Tool Room

···

From: Warren Mansell [mailto:wmansell@GMAIL.COM]
Sent: Thursday, December 19, 2013 12:22 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Classical Control Systems Lectures

Thanks Bruce! That starts to make a lot of sense, and also explains why things can get messy when I have talked to control engineers before, and also maybe explain why control theory applications to living systems got off to a difficult start with cybernetics? I will certainly watch further - sorry for my false start - as I do want to be able to communicate with real control system engineers - there are many here at the university I could be collaborating with. I am wondering though whether the human controller in the example you mention actually has a hierarchical relationship with the rheostat in controlling its reference value rather than making it open loop?

Warren

On Thu, Dec 19, 2013 at 3:23 PM, Bruce Abbott bbabbott@frontier.com wrote:

From Bruce Abbott (2014.12.19 1020 EST)

Warren Mansell –

WM: Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero? Stability within PCT relies of a flexibly changing output to ensure that the INPUT remains as close as possible to the reference value for that input? My mood is stable not because I do nothing to change it but because it remains around a range I want it to be because of, and/or despite of, my behaviour?

Warren

I agree that Douglas’ way of defining stability is unfortunate. The problem you note arises partly because he is thinking of zero as the initial value (before the system is disturbed), but doesn’t mention that explicitly. A stable system will settle to (that same) zero after being disturbed. Also contributing to the problem is that Douglas is an control systems engineer and is using an engineer’s definitions of the various quantities. The output in a standard engineering diagram of a control system is what we would label the input. For example, the output of a voltage regulator is the regulated voltage. If the reference for the voltage is +5 Volts, then the regulated output should be as close as possible to + 5 Volts.

So when Douglas asserts that in a stable system, the output goes to zero, he is not saying that the actions go to zero, but that the variable is returning to its initial state.

So, what Douglas actually is asserting is that, in a stable system, after being disturbed the variable in question will settle back to its initial value.

If you continue watching the video, you will see an illustration in the form of a ball in a valley. Its “zero” is at the bottom center of the valley. If you disturb the ball (push it uphill and release) it will eventually come to rest back at zero (assuming there’s some energy loss due to friction of viscous damping). One could define the ball’s initial position as “3 inches from the left edge of the bowl.” In that case stability would be present if the ball eventually settled at 3 inches.

It’s important to note that Douglas does recognize that what is controlled in a control system is the system’s perception of the controlled variable. In the video, Robotic Car, Closed Loop Control Example, Douglas uses a simple tracked vehicle to illustrate the difference between open-loop and closed-loop control. The open-loop version can be aimed at a particular target location and will drive there so long as there are no disturbances. But when disturbances act to turn the car, the car does not reach the target. The closed-loop version has an accelerometer that senses the angular acceleration of the car in the horizontal plane. Beginning at 8:00 minutes into the video, Douglas diagrams the system. The output on his diagram is the actual angular acceleration of the car, but Douglas notes that the only thing the control system knows about the car’s angular acceleration is the sensed value provided by the accelerometer.

In engineering, the term “control” is used in a looser sense than we use the term in PCT. A motor “controller,” for example, may consist of a simple rheostat that varies the current flow to the motor. The rheostat is said to “control” the motor speed (in the sense of being an important factor determining that speed). In the absence of any disturbances to the motor’s speed, one could make up a table or formula giving the motor’s speed for any given position of the rheostat’s dial. Such a system may be described as employing “open-loop control.” It has no ability to counteract disturbances, so a load on the motor will slow it down below the rpm given in the table.

In the diagram of this system, there’s a reference (dial setting) the inputs to the “controller” and an output, the motor’s actual speed. To convert this system to closed-loop control, you read this output with a sensor and feed the sensor’s reading (after a change in sign to negative) into a comparator. The reference moves to the comparator and the output of the comparator, the error signal, now becomes the input to the “controller.”

So-called open-loop controllers are used by human beings, who may observe the actual output and adjust the reference (e.g., the rheostat’s dial setting) to compensate for any error between what the operator wanted and that the system is actually delivering. In so doing, the operator closes the loop. The term open-loop “controller” is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I think it makes sense in that context. In PCT, of course, we prefer to restrict the term “control” to systems that compare the actual state of affairs to the desired one and take action to reduce or eliminate any difference between them.

It’s important to understand these differences in terminology (and worth the effort to learn them) so that we are not led into believing that engineers are talking nonsense about control systems when in fact they are just using somewhat different terminology and diagrams than we use in PCT when saying the same thing.

Bruce

Bruce Abbott (2014.12.18 1940 EST)

BA: For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

BA: Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control” (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture ”examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link “41 videos” next to Brian Douglas’ name just below the video, on the left to return to the page on which all the videos are displayed.

–

Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[rsmarken]

[From Rick Marken (2013.12.19.1020)]

Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero?

I see Bruce already answered you but let me put in my own 2 cents. I
liked the videos (actually I only saw the first video) that Bruce sent
because I thought it made a nice distinction between stability and
control. These are two different phenomena and confusion between the
two has caused enormous problems for PCT.

Stability is the characteristic of a variable that _returns_ to it's
initial state after a _transient_ disturbance. Thus, a ball in a bowl
is stable because after a transient disturbance the ball returns to
its initial position. Same is true of a pendulum or a mass on a
spring.

Control is a characteristic of a variable that _remains_ in a
predetermined state in the face of a _continuously varying _
disturbance.

The superficial similarity between these two phenomena is that both
seem to involve "resistance to disturbance". But the nature of this
"resistance" is quite different in the two cases. In stability, there
really is no resistance; the same forces that moved the variable away
from the initial position return it to that position once the
disturbance force is removed. In control, the resistance is active;
the resistive forces are generated by the system that is actively
acting to keep the variable in a particular state.

The problem for PCT is that psychologists who have an allergic
reaction to anything that smacks of purpose (as control does) have
assumed that the obvious controlling done by organisms is actually an
example of stability so that disturbance resistance can be explained
in terms of non-control models of stable systems. Thus, we have
"mass-spring" models of limb stability; "coordinative structure"
models of limb movement; "dynamic attractor" models of many different
activities.

I've done research to try to show that stability models cannot account
for control phenomena (Marken, R. S. (1991) Degrees of Freedom in
Behavior. Psychological Science, 2, 92 - 100). But I think part of the
problem is that psychologists don't understand (or possibly don't want
to understand) the difference between the _phenomenon_ of stability
and the _phenomenon_ of control. That's why I think the first Douglas
lecture can be useful; it can help people understand the difference
between these two phenomena and, hopefully, see that what people do is
control, not just come to rest in stable states after disturbances.

Best

Rick

···

On Thu, Dec 19, 2013 at 4:45 AM, Warren Mansell <wmansell@gmail.com> wrote:

On Thu, Dec 19, 2013 at 12:40 AM, Bruce Abbott <bbabbott@frontier.com> wrote:

From Bruce Abbott (2014.12.18 1940 EST)

For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas� lecture, �introduction to system stability and control� (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture �examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link �41 videos� next to Brian Douglas� name just below the video, on the left to return to the page on which all the videos are displayed.

Bruce A.

--
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

--
Richard S. Marken PhD
www.mindreadings.com

The only thing that will redeem mankind is cooperation.
                                                   -- Bertrand Russell

[Abbott]

Bruce Abbott (2014.12.19.1330)

Warren Mansell –

WM: Thanks Bruce! That starts to make a lot of sense, and also explains why things can get messy when I have talked to control engineers before, and also maybe explain why control theory applications to living systems got off to a difficult start with cybernetics? I will certainly watch further - sorry for my false start - as I do want to be able to communicate with real control system engineers - there are many here at the university I could be collaborating with. I am wondering though whether the human controller in the example you mention actually has a hierarchical relationship with the rheostat in controlling its reference value rather than making it open loop?

Warren

You’re welcome! As for your question, the “open-loop” controller (the rheostat) is not a negative feedback control system, so you wouldn’t really have a two-level hierarchical control system, since the bottom-level “controller” doesn’t have a comparator or feedback loop of its own. You can think of the rheostat as establishing the environmental feedback function between the operator’s output (rheostat dial setting) and the torque being generated by the motor. The torque and disturbance (load on the motor) together determine the value of the operator’s controlled variable, the motor speed.

The operator’s output function could be unpacked further as a lower-level control system whose reference is the torque to be generated by the operator’s hand on the rheostat’s knob. The reference level would be proportional to the error in perceived motor speed. The torque generated by this system would move the rheostat knob until the operator’s perceived speed of the motor matched the operator’s reference for motor speed. This more complete model of the human operator is a two-level hierarchical control system, but the rheostat is still just a device out in the operator’s environment that determines (in open-loop fashion) the relationship between rheostat dial setting and motor torque.

Bruce

p.s. The “hand-torque” lower-level system can itself be elaborated to include a yet lower-level system that sets the references for the force being generated by the muscles of the hand and arm which together produce the torque exerted on the rheostat knob. Now we’re up to a three-level hierarchy! But for most purposes we don’t need to model every level. If we assume that the lower-level control systems are doing their jobs well, we can just hide them in the person’s output function, since whatever actions the higher-level system calls far will be produced with a reasonable degree of accuracy by the lower systems. In our example, the rheostat knob will get turned (the operator’s output action) as needed to bring the motor speed to the operator’s reference motor speed.

Bruce Abbott (2014.12.19 1020 EST)

Warren Mansell –

WM: Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero? Stability within PCT relies of a flexibly changing output to ensure that the INPUT remains as close as possible to the reference value for that input? My mood is stable not because I do nothing to change it but because it remains around a range I want it to be because of, and/or despite of, my behaviour?

Warren

I agree that Douglas’ way of defining stability is unfortunate. The problem you note arises partly because he is thinking of zero as the initial value (before the system is disturbed), but doesn’t mention that explicitly. A stable system will settle to (that same) zero after being disturbed. Also contributing to the problem is that Douglas is an control systems engineer and is using an engineer’s definitions of the various quantities. The output in a standard engineering diagram of a control system is what we would label the input. For example, the output of a voltage regulator is the regulated voltage. If the reference for the voltage is +5 Volts, then the regulated output should be as close as possible to + 5 Volts.

So when Douglas asserts that in a stable system, the output goes to zero, he is not saying that the actions go to zero, but that the variable is returning to its initial state.

So, what Douglas actually is asserting is that, in a stable system, after being disturbed the variable in question will settle back to its initial value.

If you continue watching the video, you will see an illustration in the form of a ball in a valley. Its “zero” is at the bottom center of the valley. If you disturb the ball (push it uphill and release) it will eventually come to rest back at zero (assuming there’s some energy loss due to friction of viscous damping). One could define the ball’s initial position as “3 inches from the left edge of the bowl.” In that case stability would be present if the ball eventually settled at 3 inches.

It’s important to note that Douglas does recognize that what is controlled in a control system is the system’s perception of the controlled variable. In the video, Robotic Car, Closed Loop Control Example, Douglas uses a simple tracked vehicle to illustrate the difference between open-loop and closed-loop control. The open-loop version can be aimed at a particular target location and will drive there so long as there are no disturbances. But when disturbances act to turn the car, the car does not reach the target. The closed-loop version has an accelerometer that senses the angular acceleration of the car in the horizontal plane. Beginning at 8:00 minutes into the video, Douglas diagrams the system. The output on his diagram is the actual angular acceleration of the car, but Douglas notes that the only thing the control system knows about the car’s angular acceleration is the sensed value provided by the accelerometer.

In engineering, the term “control” is used in a looser sense than we use the term in PCT. A motor “controller,” for example, may consist of a simple rheostat that varies the current flow to the motor. The rheostat is said to “control” the motor speed (in the sense of being an important factor determining that speed). In the absence of any disturbances to the motor’s speed, one could make up a table or formula giving the motor’s speed for any given position of the rheostat’s dial. Such a system may be described as employing “open-loop control.” It has no ability to counteract disturbances, so a load on the motor will slow it down below the rpm given in the table.

In the diagram of this system, there’s a reference (dial setting) the inputs to the “controller” and an output, the motor’s actual speed. To convert this system to closed-loop control, you read this output with a sensor and feed the sensor’s reading (after a change in sign to negative) into a comparator. The reference moves to the comparator and the output of the comparator, the error signal, now becomes the input to the “controller.”

So-called open-loop controllers are used by human beings, who may observe the actual output and adjust the reference (e.g., the rheostat’s dial setting) to compensate for any error between what the operator wanted and that the system is actually delivering. In so doing, the operator closes the loop. The term open-loop “controller” is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I think it makes sense in that context. In PCT, of course, we prefer to restrict the term “control” to systems that compare the actual state of affairs to the desired one and take action to reduce or eliminate any difference between them.

It’s important to understand these differences in terminology (and worth the effort to learn them) so that we are not led into believing that engineers are talking nonsense about control systems when in fact they are just using somewhat different terminology and diagrams than we use in PCT when saying the same thing.

Bruce

Bruce Abbott (2014.12.18 1940 EST)

BA: For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

BA: Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control” (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture ”examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched

[wmansell]

[From Fred Nickols (2013.12.19.1243 EST)]

Consider the toys of my youth – the gunfire control systems on board UUS Navy destroyers.

The computer solves what is called the fire control problem which results in pointing a gun mount in a way that when a round is fired it will explode in the vicinity of the (moving) target.

One of the values the computer calculates is gun elevation order (i.e., the angle at which the barrel should be raised in the vertical plane). Two feedback loops come into play here.

First, there is the matter of positioning the gun elevation order module to the calculated value. The position being sent to the gun mount is by the gun elevation module is compared with the calculated value to determine if the proper gun elevation order signal is being sent.

Second, there is the matter of the gun mount’s actual position in relation to the ordered position. In this case, the receiver regulator in the gun mount compares the ordered position with the actual position.

Zero outputs indicate alignment of actual position with ordered or calculated position. BUT zero output is not what is being controlled (even though correspondence between actual and ordered is a goal). What IS being controlled is in the first case the ordered position being sent to the gun mount and in the second case the actual position of the gun mount in relation to the ordered position. In both cases what is being controlled is an input, not an output.

These gunfire control systems are designed and built by control system engineers.

Regards,

Fred Nickols, CPT

Managing Partner

Distance Consulting LLC

The Knowledge Workers’ Tool Room

···

From: Warren Mansell [mailto:wmansell@GMAIL.COM]
Sent: Thursday, December 19, 2013 12:22 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Classical Control Systems Lectures

Thanks Bruce! That starts to make a lot of sense, and also explains why things can get messy when I have talked to control engineers before, and also maybe explain why control theory applications to living systems got off to a difficult start with cybernetics? I will certainly watch further - sorry for my false start - as I do want to be able to communicate with real control system engineers - there are many here at the university I could be collaborating with. I am wondering though whether the human controller in the example you mention actually has a hierarchical relationship with the rheostat in controlling its reference value rather than making it open loop?

Warren

On Thu, Dec 19, 2013 at 3:23 PM, Bruce Abbott bbabbott@frontier.com wrote:

From Bruce Abbott (2014.12.19 1020 EST)

Warren Mansell –

WM: Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero? Stability within PCT relies of a flexibly changing output to ensure that the INPUT remains as close as possible to the reference value for that input? My mood is stable not because I do nothing to change it but because it remains around a range I want it to be because of, and/or despite of, my behaviour?

Warren

I agree that Douglas’ way of defining stability is unfortunate. The problem you note arises partly because he is thinking of zero as the initial value (before the system is disturbed), but doesn’t mention that explicitly. A stable system will settle to (that same) zero after being disturbed. Also contributing to the problem is that Douglas is an control systems engineer and is using an engineer’s definitions of the various quantities. The output in a standard engineering diagram of a control system is what we would label the input. For example, the output of a voltage regulator is the regulated voltage. If the reference for the voltage is +5 Volts, then the regulated output should be as close as possible to + 5 Volts.

So when Douglas asserts that in a stable system, the output goes to zero, he is not saying that the actions go to zero, but that the variable is returning to its initial state.

So, what Douglas actually is asserting is that, in a stable system, after being disturbed the variable in question will settle back to its initial value.

If you continue watching the video, you will see an illustration in the form of a ball in a valley. Its “zero� is at the bottom center of the valley. If you disturb the ball (push it uphill and release) it will eventually come to rest back at zero (assuming there’s some energy loss due to friction of viscous damping). One could define the ball’s initial position as “3 inches from the left edge of the bowl.� In that case stability would be present if the ball eventually settled at 3 inches.

It’s important to note that Douglas does recognize that what is controlled in a control system is the system’s perception of the controlled variable. In the video, Robotic Car, Closed Loop Control Example, Douglas uses a simple tracked vehicle to illustrate the difference between open-loop and closed-loop control. The open-loop version can be aimed at a particular target location and will drive there so long as there are no disturbances. But when disturbances act to turn the car, the car does not reach the target. The closed-loop version has an accelerometer that senses the angular acceleration of the car in the horizontal plane. Beginning at 8:00 minutes into the video, Douglas diagrams the system. The output on his diagram is the actual angular acceleration of the car, but Douglas notes that the only thing the control system knows about the car’s angular acceleration is the sensed value provided by the accelerometer.

In engineering, the term “control� is used in a looser sense than we use the term in PCT. A motor “controller,� for example, may consist of a simple rheostat that varies the current flow to the motor. The rheostat is said to “control� the motor speed (in the sense of being an important factor determining that speed). In the absence of any disturbances to the motor’s speed, one could make up a table or formula giving the motor’s speed for any given position of the rheostat’s dial. Such a system may be described as employing “open-loop control.� It has no ability to counteract disturbances, so a load on the motor will slow it down below the rpm given in the table.

In the diagram of this system, there’s a reference (dial setting) the inputs to the “controller� and an output, the motor’s actual speed. To convert this system to closed-loop control, you read this output with a sensor and feed the sensor’s reading (after a change in sign to negative) into a comparator. The reference moves to the comparator and the output of the comparator, the error signal, now becomes the input to the “controller.�

So-called open-loop controllers are used by human beings, who may observe the actual output and adjust the reference (e.g., the rheostat’s dial setting) to compensate for any error between what the operator wanted and that the system is actually delivering. In so doing, the operator closes the loop. The term open-loop “controller� is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I thtually delivering. In so doing, the operator closes the loop. The term open-loop “controller� is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I think it makes sense in that context. In PCT, of course, we prefer to restrict the term “control� to systems that compare the actual state of affairs to the desired one and take action to reduce or eliminate any difference between them.

It’s important to understand these differences in terminology (and worth the effort to learn them) so that we are not led into believing that engineers are talking nonsense about control systems when in fact they are just using somewhat different terminology and diagrams than we use in PCT when saying the same thing.

Bruce

Bruce Abbott (2014.12.18 1940 EST)

BA: For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

BA: Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control� (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture �examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link “41 videos� next to Brian Douglas’ name just below the video, on the left to return to the page on which all the videos are displayed.

–

Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[wmansell]

Thanks Rick!

···

Sent from my iPhone

On 19 Dec 2013, at 18:21, Richard Marken <rsmarken@GMAIL.COM> wrote:

[From Rick Marken (2013.12.19.1020)]

On Thu, Dec 19, 2013 at 4:45 AM, Warren Mansell <wmansell@gmail.com> wrote:

Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero?

I see Bruce already answered you but let me put in my own 2 cents. I
liked the videos (actually I only saw the first video) that Bruce sent
because I thought it made a nice distinction between stability and
control. These are two different phenomena and confusion between the
two has caused enormous problems for PCT.

Stability is the characteristic of a variable that _returns_ to it's
initial state after a _transient_ disturbance. Thus, a ball in a bowl
is stable because after a transient disturbance the ball returns to
its initial position. Same is true of a pendulum or a mass on a
spring.

Control is a characteristic of a variable that _remains_ in a
predetermined state in the face of a _continuously varying _
disturbance.

The superficial similarity between these two phenomena is that both
seem to involve "resistance to disturbance". But the nature of this
"resistance" is quite different in the two cases. In stability, there
really is no resistance; the same forces that moved the variable away
from the initial position return it to that position once the
disturbance force is removed. In control, the resistance is active;
the resistive forces are generated by the system that is actively
acting to keep the variable in a particular state.

The problem for PCT is that psychologists who have an allergic
reaction to anything that smacks of purpose (as control does) have
assumed that the obvious controlling done by organisms is actually an
example of stability so that disturbance resistance can be explained
in terms of non-control models of stable systems. Thus, we have
"mass-spring" models of limb stability; "coordinative structure"
models of limb movement; "dynamic attractor" models of many different
activities.

I've done research to try to show that stability models cannot account
for control phenomena (Marken, R. S. (1991) Degrees of Freedom in
Behavior. Psychological Science, 2, 92 - 100). But I think part of the
problem is that psychologists don't understand (or possibly don't want
to understand) the difference between the _phenomenon_ of stability
and the _phenomenon_ of control. That's why I think the first Douglas
lecture can be useful; it can help people understand the difference
between these two phenomena and, hopefully, see that what people do is
control, not just come to rest in stable states after disturbances.

Best

Rick

On Thu, Dec 19, 2013 at 12:40 AM, Bruce Abbott <bbabbott@frontier.com> wrote:

From Bruce Abbott (2014.12.18 1940 EST)

For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control�? (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture �?examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link “41 videos�? next to Brian Douglas’ name just below the video, on the left to return to the page on which all the videos are displayed.

Bruce A.

--
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

--
Richard S. Marken PhD
www.mindreadings.com

The only thing that will redeem mankind is cooperation.
                                                  -- Bertrand Russell

[wmansell]

Bruce Abbott (2014.12.19.1330)

Warren Mansell –

···

WM: Thanks Bruce! That starts to make a lot of sense, and also explains why things can get messy when I have talked to control engineers before, and also maybe explain why control theory applications to living systems got off to a difficult start with cybernetics? I will certainly watch further - sorry for my false start - as I do want to be able to communicate with real control system engineers - there are many here at the university I could be collaborating with. I am wondering though whether the human controller in the example you mention actually has a hierarchical relationship with the rheostat in controlling its reference value rather than making it open loop?

Warren

You’re welcome! As for your question, the “open-loop� controller (the rheostat) is not a negative feedback control system, so you wouldn’t really have a two-level hierarchical control system, since the bottom-level “controller� doesn’t have a comparator or feedback loop of its own. You can think of the rheostat as establishing the environmental feedback function between the operator’s output (rheostat dial setting) and the torque being generated by the motor. The torque and disturbance (load on the motor) together determine the value of the operator’s controlled variable, the motor speed.

The operator’s output function could be unpacked further as a lower-level control system whose reference is the torque to be generated by the operator’s hand on the rheostat’s knob. The reference level would be proportional to the error in perceived motor speed. The torque generated by this system would move the rheostat knob until the operator’s perceived speed of the motor matched the operator’s reference for motor speed. This more complete model of the human operator is a two-level hierarchical control system, but the rheostat is still just a device out in the operator’s environment that determines (in open-loop fashion) the relationship between rheostat dial setting and motor torque.

Bruce

p.s. The “hand-torque� lower-level system can itself be elaborated to include a yet lower-level system that sets the references for the force being generated by the muscles of the hand and arm which together produce the torque exerted on the rheostat knob. Now we’re up to a three-level hierarchy! But for most purposes we don’t need to model every level. If we assume that the lower-level control systems are doing their jobs well, we can just hide them in the person’s output function, since whatever actions the higher-level system calls far will be produced with a reasonable degree of accuracy by the lower systems. In our example, the rheostat knob will get turned (the operator’s output action) as needed to bring the motor speed to the operator’s reference motor speed.

Bruce Abbott (2014.12.19 1020 EST)

Warren Mansell –

WM: Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero? Stability within PCT relies of a flexibly changing output to ensure that the INPUT remains as close as possible to the reference value for that input? My mood is stable not because I do nothing to change it but because it remains around a range I want it to be because of, and/or despite of, my behaviour?

Warren

I agree that Douglas’ way of defining stability is unfortunate. The problem you note arises partly because he is thinking of zero as the initial value (before the system is disturbed), but doesn’t mention that explicitly. A stable system will settle to (that same) zero after being disturbed. Also contributing to the problem is that Douglas is an control systems engineer and is using an engineer’s definitions of the various quantities. The output in a standard engineering diagram of a control system is what we would label the input. For example, the output of a voltage regulator is the regulated voltage. If the reference for the voltage is +5 Volts, then the regulated output should be as close as possible to + 5 Volts.

So when Douglas asserts that in a stable system, the output goes to zero, he is not saying that the actions go to zero, but that the variable is returning to its initial state.

So, what Douglas actually is asserting is that, in a stable system, after being disturbed the variable in question will settle back to its initial value.

If you continue watching the video, you will see an illustration in the form of a ball in a valley. Its “zero� is at the bottom center of the valley. If you disturb the ball (push it uphill and release) it will eventually come to rest back at zero (assuming there’s some energy loss due to friction of viscous damping). One could define the ball’s initial position as “3 inches from the left edge of the bowl.� In that case stability would be present if the ball eventually settled at 3 inches.

It’s important to note that Douglas does recognize that what is controlled in a control system is the system’s perception of the controlled variable. In the video, Robotic Car, Closed Loop Control Example, Douglas uses a simple tracked vehicle to illustrate the difference between open-loop and closed-loop control. The open-loop version can be aimed at a particular target location and will drive there so long as there are no disturbances. But when disturbances act to turn the car, the car does not reach the target. The closed-loop version has an accelerometer that senses the angular acceleration of the car in the horizontal plane. Beginning at 8:00 minutes into the video, Douglas diagrams the system. The output on his diagram is the actual angular acceleration of the car, but Douglas notes that the only thing the control system knows about the car’s angular acceleration is the sensed value provided by the accelerometer.

In engineering, the term “control� is used in a looser sense than we use the term in PCT. A motor “controller,� for example, may consist of a simple rheostat that varies the current flow to the motor. The rheostat is said to “control� the motor speed (in the sense of being an important factor determining that speed). In the absence of any disturbances to the motor’s speed, one could make up a table or formula giving the motor’s speed for any given position of the rheostat’s dial. Such a system may be described as employing “open-loop control.� It has no ability to counteract disturbances, so a load on the motor will slow it down below the rpm given in the table.

In the diagram of this system, there’s a reference (dial setting) the inputs to the “controller� and an output, the motor’s actual speed. To convert this system to closed-loop control, you read this output with a sensor and feed the sensor’s reading (after a change in sign to negative) into a comparator. The reference moves to the comparator and the output of the comparator, the error signal, now becomes the input to the “controller.�

So-called open-loop controllers are used by human beings, who may observe the actual output and adjust the reference (e.g., the rheostat’s dial setting) to compensate for any error between what the operator wanted and that the system is actually delivering. In so doing, the operator closes the loop. The term open-loop “controller� is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I think it makes sense in that context. In PCT, of course, we prefer to restrict the term “control� to systems that compare the actual state of affairs to the desired one and take action to reduce or eliminate any difference between them.

It’s important to understand these differences in terminology (and worth the effort to learn them) so that we are not led into believing that engineers are talking nonsense about control systems when in fact they are just using somewhat different terminology and diagrams than we use in PCT when saying the same thing.

Bruce

Bruce Abbott (2014.12.18 1940 EST)

BA: For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

BA: Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control� (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture �examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched

[FredNickols]

[From Fred Nickols (2013.12.19.1431 EST)]

I’m not sure how to answer you, Warren. I was just trying to illustrate how, as an old technician, I can see a closed-loop, feedback-governed weapons system as consistent with PCT. It all depends on what you consider as an input or an output to or from what and what you assert is being controlled. In the weapons system I used one thing that is being controlled is the aiming of the gun (in the vertical and horizontal planes). Disturbances include the roll and pitch of the ship. The aiming is controlled by comparing a gun order (reference condition) with the actual position of the gun mount (perceived condition). “Upstream,� so to speak, the gun order itself is being controlled by comparing the calculated order (reference condition) with the actual position of the gun order modules (perceived condition). Does the output of the receiver regulator (which moves the gun mount) ever go to zero? Yes, theoretically speaking, when the actual position of the gun mount matches the ordered position. However, as a practical matter, it never goes to zero because the target being tracked and hence the order to the gun mount is changing. Even for a stationary target when the ship is lying to, there is the pitch and roll of the ship which causes the gun order to change in order to compensate for that movement.

I dunno if this helps or not.

Fred Nickols

···

From: Warren Mansell [mailto:wmansell@GMAIL.COM]
Sent: Thursday, December 19, 2013 2:20 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Classical Control Systems Lectures

Hi there, is the problem that in PCT we consider output as the signals sent to the effectors, whereas in control engineering they consider it as the result of the interaction between the effectors and the environment. This is the bit of the closed loop in the environment just before it enters the perceptual control system as ‘input’…

Warren

Sent from my iPhone

On 19 Dec 2013, at 17:46, Fred Nickols fred@NICKOLS.US wrote:

[From Fred Nickols (2013.12.19.1243 EST)]

Consider the toys of my youth – the gunfire control systems on board US Navy destroyers.

The computer solves what is called the fire control problem which results in pointing a gun mount in a way that when a round is fired it will explode in the vicinity of the (moving) target.

One of the values the computer calculates is gun elevation order (i.e., the angle at which the barrel should be raised in the vertical plane). Two feedback loops come into play here.

First, there is the matter of positioning the gun elevation order module to the calculated value. The position being sent to the gun mount is by the gun elevation module is compared with the calculated value to determine if the proper gun elevation order signal is being sent.

Second, there is the matter of the gun mount’s actual position in relation to the ordered position. In this case, the receiver regulator in the gun mount compares the ordered position with the actual position.

Zero outputs indicate alignment of actual position with ordered or calculated position. BUT zero output is not what is being controlled (even though correspondence between actual and ordered is a goal). What IS being controlled is in the first case the ordered position being sent to the gun mount and in the second case the actual position of the gun mount in relation to the ordered position. In both cases what is being controlled is an input, not an output.

These gunfire control systems are designed and built by control system engineers.

Regards,

Fred Nickols, CPT

Managing Partner

Distance Consulting LLC

The Knowledge Workers’ Tool Room

From: Warren Mansell [mailto:wmansell@GMAIL.COM]
Sent: Thursday, December 19, 2013 12:22 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Classical Control Systems Lectures

Thanks Bruce! That starts to make a lot of sense, and also explains why things can get messy when I have talked to control engineers before, and also maybe explain why control theory applications to living systems got off to a difficult start with cybernetics? I will certainly watch further - sorry for my false start - as I do want to be able to communicate with real control system engineers - there are many here at the university I could be collaborating with. I am wondering though whether the human controller in the example you mention actually has a hierarchical relationship with the rheostat in controlling its reference value rather than making it open loop?

Warren

On Thu, Dec 19, 2013 at 3:23 PM, Bruce Abbott bbabbott@frontier.com wrote:

From Bruce Abbott (2014.12.19 1020 EST)

Warren Mansell –

WM: Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero? Stability within PCT relies of a flexibly changing output to ensure that the INPUT remains as close as possible to the reference value for that input? My mood is stable not because I do nothing to change it but because it remains around a range I want it to be because of, and/or despite of, my behaviour?

Warren

I agree that Douglas’ way of defining stability is unfortunate. The problem you note arises partly because he is thinking of zero as the initial value (before the system is disturbed), but doesn’t mention that explicitly. A stable system will settle to (that same) zero after being disturbed. Also contributing to the problem is that Douglas is an control systems engineer and is using an engineer’s definitions of the various quantities. The output in a standard engineering diagram of a control system is what we would label the input. For example, the output of a voltage regulator is the regulated voltage. If the reference for the voltage is +5 Volts, then the regulated output should be as close as possible to + 5 Volts.

So when Douglas asserts that in a stable system, the output goes to zero, he is not saying that the actions go to zero, but that the variable is returning to its initial state.

So, what Douglas actually is asserting is that, in a stable system, after being disturbed the variable in question will settle back to its initial value.

If you continue watching the video, you will see an illustration in the form of a ball in a valley. Its “zero� is at the bottom center of the valley. If you disturb the ball (push it uphill and release) it will eventually come to rest back at zero (assuming there’s some energy loss due to friction of viscous damping). One could define the ball’s initial position as “3 inches from the left edge of the bowl.� In that case stability would be present if the ball eventually settled at 3 inches.

It’s important to note that Douglas does recognize that what is controlled in a control system is the system’s perception of the controlled variable. In the video, Robotic Car, Closed Loop Control Example, Douglas uses a simple tracked vehicle to illustrate the difference between open-loop and closed-loop control. The open-loop version can be aimed at a particular target location and will drive there so long as there are no disturbances. But when disturbances act to turn the car, the car does not reach the target. The closed-loop version has an accelerometer that senses the angular acceleration of the car in the horizontal plane. Beginning at 8:00 minutes into the video, Douglas diagrams the system. The output on his diagram is the actual angular acceleration of the car, but Douglas notes that the only thing the control system knows about the car’s angular acceleration is the sensed value provided by the accelerometer.

In engineering, the term “control� is used in a looser sense than we use the term in PCT. A motor “controller,� for example, may consist of a simple rheostat that varies the current flow to the motor. The rheostat is said to “control� the motor speed (in the sense of being an important factor determining that speed). In the absence of any disturbances to the motor’s speed, one could make up a table or formula giving the motor’s speed for any given position of the rheostat’s dial. Such a system may be described as employing “open-loop control.� It has no ability to counteract disturbances, so a load on the motor will slow it down below the rpm given in the table.

In the diagram of this system, there’s a reference (dial setting) the inputs to the “controller� and an output, the motor’s actual speed. To convert this system to closed-loop control, you read this output with a sensor and feed the sensor’s reading (after a change in sign to negative) into a comparator. The reference moves to the comparator and the output of the comparator, the error signal, now becomes the input to the “controller.�

So-called open-loop controllers are used by human beings, who may observe the actual output and adjust the reference (e.g., the rheostat’s dial setting) to compensate for any error between what the operator wanted and that the system is actually delivering. In so doing, the operator closes the loop. The term open-loop “controller� is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I thtually delivering. In so doing, the operator closes the loop. The term open-loop “controller� is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I think it makes sense in that context. In PCT, of course, we prefer to restrict the term “control� to systems that compare the actual state of affairs to the desired one and take action to reduce or eliminate any difference between them.

It’s important to understand these differences in terminology (and worth the effort to learn them) so that we are not led into believing that engineers are talking nonsense about control systems when in fact they are just using somewhat different terminology and diagrams than we use in PCT when saying the same thing.

Bruce

Bruce Abbott (2014.12.18 1940 EST)

BA: For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

BA: Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control� (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture �examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link “41 videos� next to Brian Douglas’ name just below the video, on the left to return to the page on which all the videos are displayed.

–

Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[Abbott]

[From Bruce Abbott (2013.12.19.1400 EST)]

Excellent point! But we need to be sure that stability and control are not
seen as belonging to different kinds of systems.

The system represented by the ball-in-a-bowl example is an equilibrium
system. It is stable against a transient disturbance because the energy
imparted by the transient disturbance becomes potential energy that converts
to a restorative force. In the case of the ball, the force that moves the
ball uphill against gravity creates a potential energy that is converted by
gravitational acceleration to a force that pushes the ball back to the
bottom. Similarly, compressing a spring stores energy in the spring that is
released when the spring is released, restoring the spring to its initial
length. The restoring energy comes from the disturbance.

A control system (if properly designed and tuned and operating within its
design environment) also exhibits stability, but as you note, it does so
even against a continuing disturbance. The restorative force comes, not
from the disturbance, but from the control system's own energy supply.
Consequently it can oppose a disturbance that acts continuously on the
controlled variable. (An example is keeping the biceps on your arm
contracted to keep your forearm held level.)

Both equilibrium systems and control systems can exhibit other properties
besides stability. A ball balanced on the crest a hill exhibits
instability: the slightest disturbance will send it rolling downhill. A
control system with too much gain, or too much lag, can also exhibit
instability. It may go into oscillation and even run away to infinity if
its actions get sufficiently out of phase with changes in the controlled
variable so as to convert negative feedback into positive feedback.

Bruce

Rick Marken (2013.12.19.1020) --

Hi Bruce, I switched off after the first minute. Surely, from a PCT

perspective, stability cannot be defined as the capacity to restore to an
output of zero?

I see Bruce already answered you but let me put in my own 2 cents. I liked
the videos (actually I only saw the first video) that Bruce sent because I
thought it made a nice distinction between stability and control. These are
two different phenomena and confusion between the two has caused enormous
problems for PCT.

Stability is the characteristic of a variable that _returns_ to it's initial
state after a _transient_ disturbance. Thus, a ball in a bowl is stable
because after a transient disturbance the ball returns to its initial
position. Same is true of a pendulum or a mass on a spring.

Control is a characteristic of a variable that _remains_ in a predetermined
state in the face of a _continuously varying _ disturbance.

The superficial similarity between these two phenomena is that both seem to
involve "resistance to disturbance". But the nature of this "resistance" is
quite different in the two cases. In stability, there really is no
resistance; the same forces that moved the variable away from the initial
position return it to that position once the disturbance force is removed.
In control, the resistance is active; the resistive forces are generated by
the system that is actively acting to keep the variable in a particular
state.

The problem for PCT is that psychologists who have an allergic reaction to
anything that smacks of purpose (as control does) have assumed that the
obvious controlling done by organisms is actually an example of stability so
that disturbance resistance can be explained in terms of non-control models
of stable systems. Thus, we have "mass-spring" models of limb stability;
"coordinative structure"
models of limb movement; "dynamic attractor" models of many different
activities.

I've done research to try to show that stability models cannot account for
control phenomena (Marken, R. S. (1991) Degrees of Freedom in Behavior.
Psychological Science, 2, 92 - 100). But I think part of the problem is that
psychologists don't understand (or possibly don't want to understand) the
difference between the _phenomenon_ of stability and the _phenomenon_ of
control. That's why I think the first Douglas lecture can be useful; it can
help people understand the difference between these two phenomena and,
hopefully, see that what people do is control, not just come to rest in
stable states after disturbances.

Best

Rick

From Bruce Abbott (2014.12.18 1940 EST)

For those of you who would like to learn more about the classical

approach to control systems, I highly recommend a series of YouTube videos
presented by control systems engineer Brian Douglas
(http://www.youtube.com/user/ControlLectures/videos ). The lectures range
from basic concepts to sophisticated techniques used in control system
design, analysis, and tuning, so whether you are a control systems neophyte
or an engineer who needs a bit of brushing up on these techniques, you will
find something worthwhile to view.

Douglas is an engaging and talented teacher who is able to communicate

complex ideas clearly. A nice place to start is with Douglas' lecture,
"introduction to system stability and control"
(http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture
"examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once
you have watched a given video, simply click on the link "41 videos" next to
Brian Douglas' name just below the video, on the left to return to the page
on which all the videos are displayed.

Bruce A.

--
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences Coupland I University of Manchester
Oxford Road Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for
Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT

using Method of Levels is available now.

···

On Thu, Dec 19, 2013 at 4:45 AM, Warren Mansell <wmansell@gmail.com> wrote:

On Thu, Dec 19, 2013 at 12:40 AM, Bruce Abbott <bbabbott@frontier.com> wrote:

Check www.pctweb.org for further information on Perceptual Control
Theory

--
Richard S. Marken PhD
www.mindreadings.com

The only thing that will redeem mankind is cooperation.
                                                   -- Bertrand Russell
-----
No virus found in this message.
Checked by AVG - www.avg.com
Version: 2014.0.4259 / Virus Database: 3658/6933 - Release Date: 12/19/13

[FredNickols]

[From Fred Nickols (2013.12.19.1513 EST)]

For those of you who are interested, there is what appears to be a good explanation of open-loop and closed-loop control systems (a la my Navy gun mount days) at the following link:

http://www.navymars.org/national/training/nmo_courses/NMOC/module15/14187_ch2.pdf

Fred Nickols

···

From: Fred Nickols [mailto:fred@NICKOLS.US]
Sent: Thursday, December 19, 2013 2:40 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Classical Control Systems Lectures

[From Fred Nickols (2013.12.19.1431 EST)]

I’m not sure how to answer you, Warren. I was just trying to illustrate how, as an old technician, I can see a closed-loop, feedback-governed weapons system as consistent with PCT. It all depends on what you consider as an input or an output to or from what and what you assert is being controlled. In the weapons system I used one thing that is being controlled is the aiming of the gun (in the vertical and horizontal planes). Disturbances include the roll and pitch of the ship. The aiming is controlled by comparing a gun order (reference condition) with the actual position of the gun mount (perceived condition). “Upstream,� so to speak, the gun order itself is being controlled by comparing the calculated order (reference condition) with the actual position of the gun order modules (perceived condition). Does the output of the receiver regulator (which moves the gun mount) ever go to zero? Yes, theoretically speaking, when the actual position of the gun mount matches the ordered position. However, as a practical matter, it never goes to zero because the target being tracked and hence the order to the gun mount is changing. Even for a stationary target when the ship is lying to, there is the pitch and roll of the ship which causes the gun order to change in order to compensate for that movement.

I dunno if this helps or not.

Fred Nickols

From: Warren Mansell [mailto:wmansell@GMAIL.COM]
Sent: Thursday, December 19, 2013 2:20 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Classical Control Systems Lectures

Hi there, is the problem that in PCT we consider output as the signals sent to the effectors, whereas in control engineering they consider it as the result of the interaction between the effectors and the environment. This is the bit of the closed loop in the environment just before it enters the perceptual control system as ‘input’…

Warren

Sent from my iPhone

On 19 Dec 2013, at 17:46, Fred Nickols fred@NICKOLS.US wrote:

[From Fred Nickols (2013.12.19.1243 EST)]

Consider the toys of my youth – the gunfire control systems on board US Navy destroyers.

The computer solves what is called the fire control problem which results in pointing a gun mount in a way that when a round is fired it will explode in the vicinity of the (moving) target.

One of the values the computer calculates is gun elevation order (i.e., the angle at which the barrel should be raised in the vertical plane). Two feedback loops come into play here.

First, there is the matter of positioning the gun elevation order module to the calculated value. The position being sent to the gun mount is by the gun elevation module is compared with the calculated value to determine if the proper gun elevation order signal is being sent.

Second, there is the matter of the gun mount’s actual position in relation to the ordered position. In this case, the receiver regulator in the gun mount compares the ordered position with the actual position.

Zero outputs indicate alignment of actual position with ordered or calculated position. BUT zero output is not what is being controlled (even though correspondence between actual and ordered is a goal). What IS being controlled is in the first case the ordered position being sent to the gun mount and in the second case the actual position of the gun mount in relation to the ordered position. In both cases what is being controlled is an input, not an output.

These gunfire control systems are designed and built by control system engineers.

Regards,

Fred Nickols, CPT

Managing Partner

Distance Consulting LLC

The Knowledge Workers’ Tool Room

From: Warren Mansell [mailto:wmansell@GMAIL.COM]
Sent: Thursday, December 19, 2013 12:22 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Classical Control Systems Lectures

Thanks Bruce! That starts to make a lot of sense, and also explains why things can get messy when I have talked to control engineers before, and also maybe explain why control theory applications to living systems got off to a difficult start with cybernetics? I will certainly watch further - sorry for my false start - as I do want to be able to communicate with real control system engineers - there are many here at the university I could be collaborating with. I am wondering though whether the human controller in the example you mention actually has a hierarchical relationship with the rheostat in controlling its reference value rather than making it open loop?

Warren

On Thu, Dec 19, 2013 at 3:23 PM, Bruce Abbott bbabbott@frontier.com wrote:

From Bruce Abbott (2014.12.19 1020 EST)

Warren Mansell –

WM: Hi Bruce, I switched off after the first minute. Surely, from a PCT perspective, stability cannot be defined as the capacity to restore to an output of zero? Stability within PCT relies of a flexibly changing output to ensure that the INPUT remains as close as possible to the reference value for that input? My mood is stable not because I do nothing to change it but because it remains around a range I want it to be because of, and/or despite of, my behaviour?

Warren

I agree that Douglas’ way of defining stability is unfortunate. The problem you note arises partly because he is thinking of zero as the initial value (before the system is disturbed), but doesn’t mention that explicitly. A stable system will settle to (that same) zero after being disturbed. Also contributing to the problem is that Douglas is an control systems engineer and is using an engineer’s definitions of the various quantities. The output in a standard engineering diagram of a control system is what we would label the input. For example, the output of a voltage regulator is the regulated voltage. If the reference for the voltage is +5 Volts, then the regulated output should be as close as possible to + 5 Volts.

So when Douglas asserts that in a stable system, the output goes to zero, he is not saying that the actions go to zero, but that the variable is returning to its initial state.

So, what Douglas actually is asserting is that, in a stable system, after being disturbed the variable in question will settle back to its initial value.

If you continue watching the video, you will see an illustration in the form of a ball in a valley. Its “zero� is at the bottom center of the valley. If you disturb the ball (push it uphill and release) it will eventually come to rest back at zero (assuming there’s some energy loss due to friction of viscous damping). One could define the ball’s initial position as “3 inches from the left edge of the bowl.� In that case stability would be present if the ball eventually settled at 3 inches.

It’s important to note that Douglas does recognize that what is controlled in a control system is the system’s perception of the controlled variable. In the video, Robotic Car, Closed Loop Control Example, Douglas uses a simple tracked vehicle to illustrate the difference between open-loop and closed-loop control. The open-loop version can be aimed at a particular target location and will drive there so long as there are no disturbances. But when disturbances act to turn the car, the car does not reach the target. The closed-loop version has an accelerometer that senses the angular acceleration of the car in the horizontal plane. Beginning at 8:00 minutes into the video, Douglas diagrams the system. The output on his diagram is the actual angular acceleration of the car, but Douglas notes that the only thing the control system knows about the car’s angular acceleration is the sensed value provided by the accelerometer.

In engineering, the term “control� is used in a looser sense than we use the term in PCT. A motor “controller,� for example, may consist of a simple rheostat that varies the current flow to the motor. The rheostat is said to “control� the motor speed (in the sense of being an important factor determining that speed). In the absence of any disturbances to the motor’s speed, one could make up a table or formula giving the motor’s speed for any given position of the rheostat’s dial. Such a system may be described as employing “open-loop control.� It has no ability to counteract disturbances, so a load on the motor will slow it down below the rpm given in the table.

In the diagram of this system, there’s a reference (dial setting) the inputs to the “controller� and an output, the motor’s actual speed. To convert this system to closed-loop control, you read this output with a sensor and feed the sensor’s reading (after a change in sign to negative) into a comparator. The reference moves to the comparator and the output of the comparator, the error signal, now becomes the input to the “controller.�

So-called open-loop controllers are used by human beings, who may observe the actual output and adjust the reference (e.g., the rheostat’s dial setting) to compensate for any error between what the operator wanted and that the system is actually delivering. In so doing, the operator closes the loop. The term open-loop “controller� is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I thtually delivering. In so doing, the operator closes the loop. The term open-loop “controller� is called a controller because it allows the human operator the means to control the variable in question (e.g., motor speed). I think it makes sense in that context. In PCT, of course, we prefer to restrict the term “control� to systems that compare the actual state of affairs to the desired one and take action to reduce or eliminate any difference between them.

It’s important to understand these differences in terminology (and worth the effort to learn them) so that we are not led into believing that engineers are talking nonsense about control systems when in fact they are just using somewhat different terminology and diagrams than we use in PCT when saying the same thing.

Bruce

Bruce Abbott (2014.12.18 1940 EST)

BA: For those of you who would like to learn more about the classical approach to control systems, I highly recommend a series of YouTube videos presented by control systems engineer Brian Douglas (http://www.youtube.com/user/ControlLectures/videos ). The lectures range from basic concepts to sophisticated techniques used in control system design, analysis, and tuning, so whether you are a control systems neophyte or an engineer who needs a bit of brushing up on these techniques, you will find something worthwhile to view.

BA: Douglas is an engaging and talented teacher who is able to communicate complex ideas clearly. A nice place to start is with Douglas’ lecture, “introduction to system stability and control� (http://www.youtube.com/watch?v=uqjKG32AkC4). Another is the lecture �examples of PID control (http://www.youtube.com/watch?v=XfAt6hNV8XM). Once you have watched a given video, simply click on the link “41 videos� next to Brian Douglas’ name just below the video, on the left to return to the page on which all the videos are displayed.

–

Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[Abbott]

[From Bruce Abbott (2013.12.19.1520 EST)]

Warren Mansell –

WM: Thanks again, I thought there must be a hierarchy somewhere. You know that in a hierarchy the superordinate level sets the reference value for the lower level? So in a hierarchy the reference perception being controlled at a superordinate level is of a different nature to the transformed output of that same level? So error in configuration would set references for sensations, for example? What does this mean about the lowest level if signals specifying force are sent to muscles? It might not be operating using a reference value for force. We are using a force feedback joystick that outputs force as a result of error in reference position of the joystick. And it’s working neatly at the moment…

Any thoughts people?

I wondered about that, too. Here’s what I concluded: The output of the higher-level system is just a signal. When that signal is connected to the reference of the lower-level system, it now represents the same variable, measured in the same units, as the perceptual signal to which it is compared. Thus, in the inverted pendulum demo, the output of the bob position controller connects to the reference of the bob velocity controller. Bob velocity is measured in meters/second, so the reference signal values represent meters/second.

In the bob POSITION controller, the error signal is the difference between the reference position of the bob and its perceived position, both measured in meters. So the error signal is also scaled in meters. In the output function, the error signal is multiplied by the gain: output = g*e.  The gain is usually thought of as a unitless quantity; multiplying gain times error would thus yield an output scaled in meters. But the output of this controller equals the reference for the velocity controller, which is scaled in meters/second.Â

So the gain of the position controller must serve as a conversion factor, converting so many meters of position error to so many meters/second of velocity reference.

Bruce

[MartinT]

[Martin Taylor 2013.12.19.15.35]

I think that's the wrong way to think about it. It's only to the

analyst that the signals have a different kind of significance. The
analyst sees a neural signal that varies in some way associated with
the analyst’s perception of something else in her environment. The
brain has only neural impulses, or, when a lot of them are combined
by the analyst, neural currents (and chemical variations, but we can
ignore those for now). So, the reference perception at any level is
a neural current, as is the perception to which it is compared,
nothing else. It’s not a sensation, a configuration, a sequence or
an event.
Your muscles produce forces on their environment, no matter what you
might be pushing or pulling – even if it is just other muscles.
They can’t do anything else. So the analyst looking at the muscle
system sees that the forces produced vary according to some outflow
signal (neural current). That outflow neural current is deemed by
the analyst to be a force reference signal, because whatever else
impinges on the body parts moved by the muscles, the force varies
with that output neural current. If you want to move something to a particular position, you have to
be able to perceive that position so that the neural current
corresponding to the position matches the neural current input to
what the analyst might call a position comparator. The comparator
“knows” nothing about position; nor does the perceptual function
that produces the “position” neural current. So, if you want your
joystick to produce a specified force, and to change the force it
produces you have to change the position of the stick, you have to
be able to perceive the force supplied by the joystick (by reading a
meter? by feeling the force opposing your hand?) and to be able to
have your own muscles supply forces sufficient to move the joystick
position until the force meter (or your internal force perception or
pressure on your skin) reads what you want it to read. Its difficult to remember that none of the signals in the brain are
what the analyst might say they represent. They are neural current
values, or spike rates. or polarization waveforms, and they are the
same at all levels in the hierarchy and in all the functions of
every control loop. The only places where this is untrue are in the
environment outside the skin and at the sensor and muscular
interfaces to the environment. There, you can use units like
metres/sec, kg/m^3, and so forth. Internally, the only units are
spikes/sec (or equivalent).
Does that resolve the dilemma?
Martin

···

    Thanks again, I thought there must be a hierarchy somewhere.

You know that in a hierarchy the superordinate level sets the
reference value for the lower level? So in a hierarchy the
reference perception being controlled at a superordinate level
is of a different nature to the transformed output of that same
level?

    So error in configuration would set references for

sensations, for example? What does this mean about the lowest
level if signals specifying force are sent to muscles? It might
not be operating using a reference value for force. We are using
a force feedback joystick that outputs force as a result of
error in reference position of the joystick. And it’s working
neatly at the moment…

Any thoughts people?

Warren

[wmansell]

Exactly, and if for every other level of the hierarchy, the output signal acting as the reference for the lower level is always a different code / perception, could this not be going on at the lowest level too? So, no force needs to be sensed?
Warren

···

On Thu, Dec 19, 2013 at 8:20 PM, Bruce Abbott bbabbott@frontier.com wrote:

[From Bruce Abbott (2013.12.19.1520 EST)]

Warren Mansell –

WM: Thanks again, I thought there must be a hierarchy somewhere. You know that in a hierarchy the superordinate level sets the reference value for the lower level? So in a hierarchy the reference perception being controlled at a superordinate level is of a different nature to the transformed output of that same level? So error in configuration would set references for sensations, for example? What does this mean about the lowest level if signals specifying force are sent to muscles? It might not be operating using a reference value for force. We are using a force feedback joystick that outputs force as a result of error in reference position of the joystick. And it’s working neatly at the moment…

Any thoughts people?

I wondered about that, too. Here’s what I concluded: The output of the higher-level system is just a signal. When that signal is connected to the reference of the lower-level system, it now represents the same variable, measured in the same units, as the perceptual signal to which it is compared. Thus, in the inverted pendulum demo, the output of the bob position controller connects to the reference of the bob velocity controller. Bob velocity is measured in meters/second, so the reference signal values represent meters/second.

In the bob POSITION controller, the error signal is the difference between the reference position of the bob and its perceived position, both measured in meters. So the error signal is also scaled in meters. In the output function, the error signal is multiplied by the gain: output = g*e. The gain is usually thought of as a unitless quantity; multiplying gain times error would thus yield an output scaled in meters. But the output of this controller equals the reference for the velocity controller, which is scaled in meters/second.

So the gain of the position controller must serve as a conversion factor, converting so many meters of position error to so many meters/second of velocity reference.

Bruce

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist

School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[wmansell]

Ah, I see with Martin’s comments now - the lowest level sends out a neural signal for the muscle that represents its desired force at that moment, but then the muscle needs to do whatever it does at an anatomical/physical level to ensure that this reference signal of force is produced, is that right?
Warren

···

On Fri, Dec 20, 2013 at 11:54 AM, Warren Mansell wmansell@gmail.com wrote:

Warren

Exactly, and if for every other level of the hierarchy, the output signal acting as the reference for the lower level is always a different code / perception, could this not be going on at the lowest level too? So, no force needs to be sensed?

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

On Thu, Dec 19, 2013 at 8:20 PM, Bruce Abbott bbabbott@frontier.com wrote:

[From Bruce Abbott (2013.12.19.1520 EST)]

Warren Mansell –

WM: Thanks again, I thought there must be a hierarchy somewhere. You know that in a hierarchy the superordinate level sets the reference value for the lower level? So in a hierarchy the reference perception being controlled at a superordinate level is of a different nature to the transformed output of that same level? So error in configuration would set references for sensations, for example? What does this mean about the lowest level if signals specifying force are sent to muscles? It might not be operating using a reference value for force. We are using a force feedback joystick that outputs force as a result of error in reference position of the joystick. And it’s working neatly at the moment…

Any thoughts people?

I wondered about that, too. Here’s what I concluded: The output of the higher-level system is just a signal. When that signal is connected to the reference of the lower-level system, it now represents the same variable, measured in the same units, as the perceptual signal to which it is compared. Thus, in the inverted pendulum demo, the output of the bob position controller connects to the reference of the bob velocity controller. Bob velocity is measured in meters/second, so the reference signal values represent meters/second.

In the bob POSITION controller, the error signal is the difference between the reference position of the bob and its perceived position, both measured in meters. So the error signal is also scaled in meters. In the output function, the error signal is multiplied by the gain: output = g*e. The gain is usually thought of as a unitless quantity; multiplying gain times error would thus yield an output scaled in meters. But the output of this controller equals the reference for the velocity controller, which is scaled in meters/second.

So the gain of the position controller must serve as a conversion factor, converting so many meters of position error to so many meters/second of velocity reference.

Bruce

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist

School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[wmansell]

Maybe we need to look at the circuitry of our joystick to see if it used a negative feedback loop to check that it is applying the force that the computer tells it to? I can’t quite work it out, but it certain shows that for the joystick there are some very detailed mechanics going on beyond the point at which it receives its signal to elicit a force. This does make me think that we need to accept that the biological and physical dynamics of the body within our PCT model will be very important, but nonetheless still could involve a form of negative feedback?

http://www.google.co.uk/patents?hl=en&lr=&vid=USPAT5742278&id=5D0jAAAAEBAJ&oi=fnd&dq=force+feedback+joystick+patent&printsec=abstract#v=onepage&q=force%20feedback%20joystick%20patent&f=false

···

On Fri, Dec 20, 2013 at 11:56 AM, Warren Mansell wmansell@gmail.com wrote:

Warren

Ah, I see with Martin’s comments now - the lowest level sends out a neural signal for the muscle that represents its desired force at that moment, but then the muscle needs to do whatever it does at an anatomical/physical level to ensure that this reference signal of force is produced, is that right?

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

On Fri, Dec 20, 2013 at 11:54 AM, Warren Mansell wmansell@gmail.com wrote:

Warren

Exactly, and if for every other level of the hierarchy, the output signal acting as the reference for the lower level is always a different code / perception, could this not be going on at the lowest level too? So, no force needs to be sensed?

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

On Thu, Dec 19, 2013 at 8:20 PM, Bruce Abbott bbabbott@frontier.com wrote:

[From Bruce Abbott (2013.12.19.1520 EST)]

Warren Mansell –

WM: Thanks again, I thought there must be a hierarchy somewhere. You know that in a hierarchy the superordinate level sets the reference value for the lower level? So in a hierarchy the reference perception being controlled at a superordinate level is of a different nature to the transformed output of that same level? So error in configuration would set references for sensations, for example? What does this mean about the lowest level if signals specifying force are sent to muscles? It might not be operating using a reference value for force. We are using a force feedback joystick that outputs force as a result of error in reference position of the joystick. And it’s working neatly at the moment…

Any thoughts people?

I wondered about that, too. Here’s what I concluded: The output of the higher-level system is just a signal. When that signal is connected to the reference of the lower-level system, it now represents the same variable, measured in the same units, as the perceptual signal to which it is compared. Thus, in the inverted pendulum demo, the output of the bob position controller connects to the reference of the bob velocity controller. Bob velocity is measured in meters/second, so the reference signal values represent meters/second.

In the bob POSITION controller, the error signal is the difference between the reference position of the bob and its perceived position, both measured in meters. So the error signal is also scaled in meters. In the output function, the error signal is multiplied by the gain: output = g*e. The gain is usually thought of as a unitless quantity; multiplying gain times error would thus yield an output scaled in meters. But the output of this controller equals the reference for the velocity controller, which is scaled in meters/second.

So the gain of the position controller must serve as a conversion factor, converting so many meters of position error to so many meters/second of velocity reference.

Bruce

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist

School of Psychological Sciences
Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[Abbott]

[From Bruce Abbott (2013.12.20.0905 EST)]

Warren Mansell –

WM: Maybe we need to look at the circuitry of our joystick to see if it used a negative feedback loop to check that it is applying the force that the computer tells it to? I can’t quite work it out, but it certain shows that for the joystick there are some very detailed mechanics going on beyond the point at which it receives its signal to elicit a force. This does make me think that we need to accept that the biological and physical dynamics of the body within our PCT model will be very important, but nonetheless still could involve a form of negative feedback?

http://www.google.co.uk/patents?hl=en&lr=&vid=USPAT5742278&id=5D0jAAAAEBAJ&oi=fnd&dq=force+feedback+joystick+patent&printsec=abstract#v=onepage&q=force%20feedback%20joystick%20patent&f=false

I looked at the patent briefly and did not see any evidence that negative feedback control is employed by the joystick circuitry. Such control would treat the user’s attempts to move the joystick as disturbances and the system would apply whatever force is necessary (within the limits of its ability to generate force) to prevent the user’s actions from changing the controlled variable – which is not the way you’d want the joystick to behave. Instead, what you want is for the joystick to generate forces that summate with the forces that the user inputs to the stick. For example, the device can simulate a spring-loaded joystick in which a restoring force is generated that pushes the stick back toward the center position whenever the user moves the stick away from center. The strength of force would increase with the deviation from center. The device can also apply a damping force proportional to the rate at which the stick is moved. In this case the stick would feel like it was moving through a viscous liquid. The computer to which the stick connects sends instructions to the processor onboard the joystick telling it how to behave – what forces to generate relative to how and where the stick is moved; the onboard processor senses the stick’s position and from change in position per unit of time computes the velocity of movement, and then applies forces according to these values (depending on what kind of effect or effects have been selected) in open-loop fashion.

You are correct that the biological and physical dynamics of the body will be very important – when an attempt is made to create a detailed and highly accurate model of these systems. In our current PCT simulations, more programming is devoted to creating a simulation of the physical environment with which the person will interact than to simulate the relevant control systems within the individual. Fortunately, we can model many control situations without having to represent those control systems in great detail. Most of how a control system behaves is due to the organization of its parts rather to the details of how those parts are constructed, and for many purposes we can just assume that the lower-level systems are doing their jobs and “hide” them inside the output functions of the systems being modeled.

Bruce