Defining control. Review of literature
Unedited posts from archives of CSG-L (see INTROCSG.NET):
Date: Mon, 29 May 1995 14:31:38 -0600
Subject: Defining control
[From Bill Powers (950529.1410 MDT)]
It seems that all we have to do to distinguish the PCT meaning of controlfrom other meanings is to call it "negative feedback control." My referencetexts aren't very up-to-date,but this is what I found in two of them:
Phillips, C. L. and Harbor, R. D. (1988). _Feedback controlsystems_.
To control any physical variable, which we usually call a _signal_, wemust know the value of this variable, that is, we must measure this variable.We call the system for the measurement of this variable a _sensor_... We definethe _plant_ of a control system as that part of the environment to becontrolled. (p. 1)
Ogata, K. (1970) _Modern Control Engineering_.
FEEDBACK CONTROL. Feedback control is an operation which, in the presenceof disturbances, tends to reduce the difference between the output of a systemand the reference input (or an arbitrarily varied, desired state) and whichdoes so on the basis of this difference. Here, only unpredictable disturbances(i.e., those unknown beforehand) are designated for as such, since withpredictable or known disturbances, it is always possible to includecompensation within the system so that measurements are unnecessary.
Maybe some others can look up more definitions.
Best, Bill P.
Date: Tue, 30 May 1995 02:46:48 -0400
Subject: Re: Defining control
<[Bill Leach 950530.02:03 U.S. Eastern Time Zone]
>[From Bill Powers (950529.1410 MDT)]
To my shock and surprise, I only found one reference that properly definedthe term "control". The McGraw-HillElectronics Engineers Handbook does have a section entitled "Control Theory"and presents the basic control loop in diagrammatical form with which we areall so familiar (including proper labeling).
All of the electronic "dictionaries" were abysmal in their treatment of theterm "control" or "control system".
The physics books and Dictionary for Science Writers that I have wereequally abysmal.
Unfortunately, even the term "negative feedback" is treated poorly for ourpurposes. The problem of course is that for an amplifier that is "stabilized"with negative feedback IT IS the actual value of the output signal itself thatis "desired" to match the input times some transform and it is the feedback(with appropriate inverse function) and almost unlimited "open loop" gain thatproduces this result.
Additionally, negative feedback in amplifiers is not nearly as straightforward in principle as what we are trying to deal with. That is, we don'tconcern ourselves with feedback within a single stage or between a few stagesas do the electronics folks. Also we are not dealing with "cathode" feedbackand the like where the feedback signal has a significantly different affectupon the circuit as the reference on the grid. If the "electronics" definitionof feedback dealt only with Operational Amplifiers then their definition wouldlook much more like ours.
-bill
Date: Thu, 1 Jun 1995 18:00:38 -0400
Subject: Definitions
<[Bill Leach 950601.14:34]
>NET/Bill Powers (request)
My comments (if any) are in []. Leading & trailing underscore impliesemphasized text in the source.
From "Electronics Learning Dictionary" Howard W. Sams:
Control
Also called a control circuit.
1. In a digital computer ...
2. Sometimes called manual control. In any mechanism, one or morecomponents responsible for interpreting and carrying out manually initiateddirections. [this is certainly gives as a great deal of concise and unambiguousinformation --lets see now, control is ahhh]
3. In some business applications, a mathematical check.
4. In electronics, a potentiometer or variable resistor.
5. In an alarm system, any mechanism which sequences the interrogation ofprotected site units, resets latched alarms and performs similarfunctions.
Controller
1. An instrument that holds a process or condition at a desired level orstatus as determined by comparison of the actual value with the desired. [nottoo bad but unfortunately the use of the word "actual" for the perceptiondoubtlessly is a source for misunderstanding]
2. A device of group of devices, which serves to govern in somepredetermined manner, the electric power delivered to the apparatus to which itis connected. [this is of course the "component" called a "controller"]
Feedback
1. In a transmission system or a section of it, the returning of a fractionof the output to the input.
2. In a magnetic amplifier, a circuit connection for which no additionalmagnetomotive force (which is a function of the quantity) is used to influencethe operating conditions. [Yes I DID recheck my typing on this one! My My thatis profound. ... influence the operating cond.]
3. In a control system, the signal or signals returned from a controlledprocess to denote its response to the command signal. Feedback is derived froma comparison of actual response to desired response, and any variation is usedas an error signal combined with the original control signal to help attainproper system operation. Systems employing feedback are termed closed-loopsystems; feedback closes the loop. [this is really pretty good except again forthe assumption that the perception is the actual value as opposed to the sensoroutput]
4. Squeal or howl from speaker caused by speaker sound entering microphoneof a recorder or amplifier.
5. The return of a portion of the output of a circuit or device to itsinput. With positive feedback, the signal fed back is in phase with the inputand increases amplification, but may cause oscillation. With negative feedback,the signal is 180 degrees out of phase with the input and decreasesamplification but stabilizes circuit performance and tends to lower anamplifier's output impedance, improve signal stability and minimize noise anddistortion.
Feedback Control
1. A type of system control obtained when a portion of the output signal isoperated upon and fed back to the input in order to obtain a desired effect.[another completely unambiguous explanation!]
2. An automatic means of sensing speed variation and correcting to maintaina constant speed or close speed regulation.
Feedback Control Loop
A closed transmission path which includes an active transducer andconsists of a forward path, a feedback path, and one or more mixing pointsarranged to maintain a prescribed relationship between the loop input andoutput signals.
Feedback Control Signal
That portion of the output signal which is returned to the input in orderto achieve a desired effect, such as fast response.
Feedback Control System
A control system comprising one or more feedback control loops; itcombines the functions of the controlled signals and commands, tending tomaintain a prescribed relationship between the two.
From "Dictionary of Electronics" Funk & Wagnalls
Control
*No Entry* [believe it or not]
Feedback
The return of a part of a system output to the system input, causing, ingeneral, a profound change in the characteristics of the system. [can't say asI know a whole lot more than before reading this] If the returned signal is inphase with the input, it is called _positive_ or _regenerative feedback_; if itis out of phase it is called _negative_ or _degenerative feedback_. Negativefeedback improves the stability and linearity of a system at the expense ofGAIN, while positive feedback increases gain and speed of response but makesthe system less stable and more oscillatory. Consider ... [of course thisdescription is not considering control system design where the open loop gainof the output function is often 1e7 or greater]
Feedback Control System
A system containing a set of devices that measure its output or outputsagainst an appropriate set of reference signals, generating a set of errorsignals which in turn control the system in such a way that its outputs conformto desired performance criteria. The performance, that is, the ratio betweenoutput c and reference r, of the generalized system shown is given by
c/r=KHG/(1+KFHG).
The system becomes unstable if 1+KFGH=0, that is, if the phase shiftaround the loop H-K-G-Fis zero when KHFG=1. See FEEDBACK
----- ----- -----
/-\ | | | | | |
--->SUM--->|H |--->|K |--->|G |------>
\-/ | | | | | | |
^ ----- ----- ----- |
| Series amp Ctrlled |
| comp. Plant |
| ----- |
| | | |
|-----------|F |---------------|
| |
-----
Feedback
Compensator
From "Oxford Dictionary for Scientific Writers and Editors" OxfordPress
Control and Feedback both have entries without definition.
From "Control Technology and Personal Computers" Van NostrandReinhold
[A disappointment in the extreme but no where in scanning the book did Ifind a single example of the use of the term "control" such that one couldconclude that the author had even the vaguest idea of what a control system isand what it is supposed to be doing.]
From "McGraw-HillEncyclopedia of Science & Technology" [guess]
[And if you have ever wondered why government "control" of the economy ismost generally a disaster... here is the model:]
private business
investment
|
|
gvt spending-----| |
| |+
----------v v ------------
desired /-\ | | + /-\ | business |
national-->|SUM|-->|government|--->|SUM|--->|production|---->
income \-/ | | \-/ | | |
^ ------------ ^ ------------ |nat'l
| +| |income
| | consumer |
| | spending |
| | |
-------------- ------------- |
| | | | |
| management | | consumers | |
| | | | |
-------------- ------------- |
^ ^ |
| | |
| | |
| | --------------- |
| /-\-| tax | |
| |SUM|<-|collections |<--|
| \-/ | | |
| ^ --------------- |
| |+ |
| | |
------------------------|-----------------------
[Actually the presentation in this book set looks to be decent (hundreds ofpages devoted to various control system issues).]
From "Encyclopedia of Physics" VCH Pub.
Servomechanism
...
The basic configuration of a servomechanism is shown by the solid lines inFig. 1. [not included as it IS the basic control diag] A position controlsystem may be required to drive the output, C, to some commanded position R.It may also be required to act as a regulator and keep the output C at thedesired R despite the application of some load disturbance D (a typical loaddisturbance is a wind load on a radar antenna). A basic requirement is thatthe output C be measured at least as accurately as required by the positioningspecifications. [not bad at all] ...
[The entire article is quite good as it is intended to be descriptive asopposed to design information and yet "hits" the issues well]
I believe someone once said something along the lines of "Be careful ofwhat you ask for... you just might get it!"
-bill
Date: Thu, 1 Jun 1995 18:01:12 -0400
Subject: Thought I was done Huh?
<[Bill Leach 950601.14:35] >NET
[Bet y'all thought that last missive was all of it huh?]
From "Electronic Engineer's Handbook" McGraw-Hill
[This book contains the best and most complete control system theorytreatment of the books that I have. ie: There is more on Kalman filters thanany of us would ever want to see!]
Modeling
By and large, models constitute the realm of discourse within which systemengineering is carried on. More pragmatically, a model is a prerequisite to theuse of analytical methods in engineering design.
By _modeling_ we mean any deliberate intelligible cognitive activity aimedat abstracting, and reproducing in some convenient realm of discourse, featuresof an object or system (the prototype) of interest to the modeler. The activityis deemed cognitive ...
Of late, models are becoming generally recognized as indispensable toolsfor effective understanding of the behavior of complex systems. Yetmathematical modeling is still, at best, an art. There is no comprehensive,consistent body of theory which constitutes a theory of modeling. ...
Engineering systems modeling is a blending of physical and mathematicaltheory. It is a sterile activity if either is left out. In the sciences, modelsare sought which illuminate natural phenomena. The objective is to strip awayall that is not essential so that our observations of reality can becharacterized and understood in terms of some ultimate simplicity. ... In thiscontext a model is a _THEORY_ constituting a set of propositions of laws fromwhich facts exhibited in nature can be deduced.
This last notion illustrates what has come to be called the _scientificmethod_. The scientific method of establishing an understanding of any physicalphenomenon is generally identified as consisting of three phases (1) _initialobservation_, (2) _formulation of a theory_, and (3) prediction of newobservations and experimentation. Moreover, the completion of the last stagefrequently suggests refinements ... The emphasis on observation has its rootsin the empiricist philosophy which has been at the heart of modernscience.
...
One useful classification of models distinguishes three types: (1) _nativemodels_ (the past trends of a single variable are used to predict futurebehavior of that variable); (2) _simple correlative models (past observationsare used to correlate several interrelated variables in order to forecastfuture trends); and (3) _causal models_ (the response of certain variables dueto changes in others is predicted).
...
A fundamental distinction can be drawn between the first two model typesand causal models. Naive and simple correlative models are _descriptive_,whereas causal models are _explanatory_. ...
...
...
...
Compared with ontological issue of parsimony, an even thornier issue theepistemological problem of model _validation_. Without a clear understanding ofthe relationship between a model and its prototype, it is not clear hownecessary and sufficient conditions for validation can be established (or evenhow "validation" can be defined unambiguously). The problem of how a model,e.g., a mathematical system, relates to its prototype, e.g. a physical or asocial system, is rarely addressed. In consequence, discussions of validationare diverse and inconclusive. ...
CONTROL THEORY
In Par. 5-1we asserted that systems engineering deals with the understanding of system assuch, for which an understanding of the components is necessary but notsufficient. From this point of view, perhaps the epitome of systems engineeringis the sense of an exemplar of archetype, is _control theory_, and the essenceof control theory is found in the concept of _feedback_.
...
...
...
[And Bill P. will truly _love_ this next one]
Compensation
The basic idea of feedback is intuitive and simple. From the perspectiveof a human operator attempting any control action, whether that of positioninga lamp on a table, steering an automobile, or any of the innumerable actions wetake continually and instinctively, our action is almost invariably tempered byour continuing observation of any discrepancy between intent and status thusfar. [The next one however is in serious error unless the author meant"results" when he said "output"] This is negative feedback: The control actionis a function of the difference between the desired output and the actualoutput.
From "Electronic Engineer's Reference Book" Butterworths
[From a descriptive standpoint there is almost nothing worth quoting fromthis book however, the diagram for "closed loop control" is worthreproducing:]
Summing
junction ____________ __________ _________
/-\ | | | | | |
0i ----->|X|-->|Controller|->|Actuator|->|Process|------>0o
\-/ | | | | | | |
Reference ^ ------------ ---------- --------- |
signal | |
|__________________________________________|
Signal from
sensor
-bill
Date: Thu, 1 Jun 1995 18:09:47 -0500
Subject: Recent Definitions
[From Bruce Abbott (950601.1805 EST)]
>Bill Powers (950529.1410 MDT) --
> Maybe some others can look up more definitions [of control].
I keep forgetting to bring my text on designing linear control systems into work; its definitions are similar to those you quote from Phillips andHarbor (1988). But here's some interesting definitions in Chapter 3 (ControlTheory Overview) of a 1990 text in reference to controlling a motor:
"To control" is defined as "to manipulate an object (motor) so as to servea certain purpose" (so as to make it work as required).
In the aforementioned case, the motor is called the controlled system. Thephysical quantities (voltage, current, frequency, torque, angular velocity, andangle) are called the controlled variables, the required manipulations arecalled the controls (voltage, current, and frequency), and the instruction tobe given is called the control reference.
Control is classified into feedback control and feedforward control.Feedback control detects the controlled variables, then compares them with thecontrol reference to determine the control variables.
Feedback control is insensitive to disturbances (load torque fluctuation,source voltage fluctuation, etc.) which disturb the behavior of the system, andto the parameter variation (change in inertia, resistance, etc.). Feedbackcontrol also changes the structure of control systems.
Feedforward control makes the response of the control system fasterbecause it determines the controls using future information. Example: when youdrive a car on the road, you will see what the road ahead looks like. Thecontrol reference is the desired value or input, and the controlled variable iscalled the output. The difference between the desired value and the controlledvariable is called error.
Dote, Yasuhiko. (1990). _Servo motor and motion control using digitalsignal processors. Englewood Cliffs, NJ: Prentice Hall.
There is quite a bit more I could add but this is probably enough forpresent purposes. Early in the book the word "plant" is used to refer to themotor but "plant" does not appear in the index and there is no glossary.Published by Prentice Hall, book is part of a series sponsored by TexasInstruments.
Many years ago (when I was in high school) I had an idea for an automobilesuspension system in which some kind of sensor would monitor the distance fromthe chassis to the road surface (independently at each wheel). This was to bemounted immediately IN FRONT of the wheel. If a bump in the road went underthe sensor, then the system, acting by means of a hydraulic piston/shock, wouldbegin to pull the wheel up just as the bump began to pass under the wheel. Solong as there was sufficient travel, the force supporting the car would remainconstant and there would be no chassis movement in the vertical plane. Thereverse action would occur for dips or potholes. By virtue of its sensorposition, this device would have functioned something like the feedforwardcontrol system described above. I thought the lead-timewould be needed to compensate for system inertia.
Regards, Bruce
Date: Fri, 2 Jun 1995 02:08:44 -0400
Subject: Re: Recent Definitions
<[Bill Leach 950602.01:35]
>[From Bruce Abbott (950601.1805 EST)]
I really should not have "pulled down" that next batch ofmessages...
>> Feedback control is insensitive to disturbances (load torquefluctuation, source voltage fluctuation, etc.) which disturb the behavior ofthe system, and to the parameter variation (change in inertia, resistance,etc.). Feedback control also changes the structure of control systems.
This is absolutely false based upon the listing you gave of the controlledvariables. A closed loop negative feedback system monitoring the variables youlisted and changing the voltage, current and frequency WILL keep the motorrunning at the desired conditions within the physical capability of the motorand the available power (ie: I am presuming that the controller can deliverfull overload current to the motor without itself failing).
The feedforward that they are probably referring to in that article (atleast with respect to the motor controller) is a derivative feedback loop (orconversely a rate sensitive output device, that is an amplifier that increasesits' gain as a function of the rate of change of input).
Another aspect of control that is sometimes used and generally consideredto be "feedforward" control (and in my mind is actually the only REALfeedforward control that can exist) is that large industrial motor controllers(usually several thousands of horsepower) that also have very large percentage"step" change in load but still must control to a close tolerance will oftenhave an additional input to the controller from one or more of the load devicesto indicate that the load is being applied or removed.
This is really a "model" design aspect and both the timing and magnitude ofeffect have to be "tuned" in actual operation (though digital systems will nowdo the tuning on their own --ie: learning).
> cars suspension
This is "doable" but you don't really gain much for the effort since thewheel so moved must be restored to its original relationship with the chassis --not to mention that you would probably have to haul the hydraulic plant in atrailer if the car was to remain smaller than the Enterprise!
-bill
Date: Fri, 2 Jun 1995 07:46:14 -0600
Subject: Re: PCT research
[From Bill Powers (950602.0600 MDT)]
Bill Leach (950601.14:34 )--
Great! Thanks for the many horrible examples of defining control and thefew good ones. I'd particularly like to see what engineers are being taught,and have been taught in the last 20 years or so. All you have to do is look atthe first few pages of textbooks. In my experience, after an extremely briefintroduction textbooks plunge right into transform methods or othermathematical treatments that give essentially no feel for how control systemsactually work.
The McGraw-Hill"Electronic Engineer's Handbook" entries were great.
> [The next one however is in serious error unless the author meant"results" when he said "output"] This is negative feedback: The control actionis a function of the difference between the desired output and the actualoutput.
In engineering parlance, the "output" is the controlled variable. The linkbetween the physical effector and the measure of the output is often lumpedinto the effector unless there is something, like a long shaft subject totorsion, between the effector and the "output" that is desired to becontrolled. Engineers, for obvious reasons, like to keep the link between theprimary effector and the controlled variable as short and simple as possible,which may be why some of them have trouble understanding the idea of control of_input_. For them, the controlled variable is most closely associated with theeffector output, and the sensors are just the means of getting a feedbacksignal.
----------------------------------------
Bruce Abbott (950601.1805 EST) --
More good (horrible) stuff from the literature!
> "To control" is defined as "to manipulate an object (motor) so as toserve a certain purpose" (so as to make it work as required).
This is a very popular one: Francisco Varela uses it. It fails to mentionWHOSE purpose is involved, and how using something for a purpose works.Basically, it's the output blunder: confusing the means of control with controlitself.
Neat idea for a bump-suppressor.The lead of the sensor would indeed make the control problem simpler -but you'd have to be able to move the bump sensor ahead and back as a functionof speed. Also, you'd have to limit the frequency response, to avoid hittingthe stops when going over a long bump or dip. If you could design the system touse minimum energy (storing most of the energy in springs), this might be veryfeasible.
-------------------
Bill Leach (subsequent posts)--
BILL! GO TO BED!
Best to all, Bill P.
Date: Sat, 3 Jun 1995 15:38:27 -0500
Subject: Re: Defining control
[from Bruce Buchanan 950603:1540 EDT]
As an occasional lurker I have noted Bill Powers promotion of discussioninter alia of terminology related to Control and Negative Feedback...
A number of definitions have been cited. The problem with many of them isthat they are of a technical nature and derivation, not at a level of generalprinciple which carries well to other fields or has potentially largerimplications...
So here are some thoughts - in the spirit of Bill Leach [950601.14:34 EDT] -(per the Sorcerer to the Apprentice?)
> I believe someone once said something along the lines of "Be carefulof what you ask for... you just might get it!"
An article in the June/95 Scientific American ("From Complexity toPerplexity") suggests that the current approaches to questions of complexity,which include systems of dynamic control, are far from successful. So adiscussion of some fundamentals may not be out of place.
I take it that a principle objective of science is the creation of abstractmodels valid for classes of systems. For example, physics, at its highestlevel, deals with matter as energy, manifested fundamentally in heat and playedout through variously organized structures in the photosynthetic, oxidative andmetabolic processes which support life. One does not require mathematicalprecision to understand this much.
In describing these things, we of course discuss conceptual models ofreality, not Boss reality itself. Nevertheless, such processes must involveorganization in space and time -presumably in the world and certainly in the concepts. And any models orpatterns of which we can speak must also reflect information and someselectivity or choice. So what are the irreducible fundamentals of ourinformation about things?
The thesis I have in mind (not as original, but for emphasis anddiscussion) is that the most basic and primary level is the cybernetic model ofnegative feedback. Below this level of organization events do not exist for us(unknowable because unrepeated), there are no identifiable elements, nopossibility of patterned relationships to be conceived. Events and elementsbecome identifiable insofar as they participate and can be accessed repeatedlyand via feedback processes, which are essential to any structure ofrelationships involving time, hence any perceivable or knowable data. (I amprobably repeating thing Bill Powers has written somewhere.)
Considerations:
An essential primary process/condition for any existent or existence itselfis some persistence through time. It has been believed, and still is believedby many, that the external world consisted of objects, e.g. rocks and peopleand 'facts', and certainly we can perceive the world in this way. Morediscriminating or selective perception reveals, however, that underlying andwithin primary experience are dynamic processes, including perceptualprocesses, in constant renewal and interaction, both in the world and withinthe organism. Persistence and renewal through time requires the action ofcircular causal processes through which the enduring world and therepeatability and consistent identification of perceptions becomepossible.
In relation to the external world we speak of causal processes, which areselections we make both among (1) interactive relations among events as theseplay out in time (which in engineering may include Feedforward), and (2)retroactive or circular causal loops which link elements and events to repeatpatterns in time, dampened depending upon Negative Feedback, but always withsome delay (e.g. hysteresis) which is required and provides the existentialcondition for possibilities for adaptive change. The relation with time iscrucial. Disturbance must actually occur, then be organized via negativefeedback, for random contingencies to be overcome, and for life to arise andendure.
Examples of relatively stable cyclic processes in the inorganic world,which also impact upon other events and life in changing ways, are thehydrologic cycle, the magnetosphere, jet stream, etc. etc. For negativefeedback also underlies the stability of universal processes.
In relation to man's thought and the conceptual world e.g. science, it maybe recognized that sensation and perception occur over many levels which areall, in principle, subject to control by feedback in terms of variouscriteria.
Perhaps the most accurate overall view of this process is that whichrecognizes that it is the activity or behavior of the organism which controlsperception (cf. B:CP). (I am just trying to paint the picture as I see it, inorder to clarify ideas, not preach to the converted!)
External stimuli may be perceived, but they do not control, since they arenot as such properties of the system. Subsequent events and memory may changethis, and the picture may be confused if the chronology is notrecognized.
So there are other views ( e.g. S-R)which see behavior as determined simply by What is perceived. Such views areemployed because they appear to the proponents be useful for special purposes,and may be validated by repetition within tightly constrained conditions.However, they do not take into account the actual nature of the fundamentaldynamic processes at work, and do not meet criteria for science that reach forvalid formulations which are the most comprehensive.
Terminology:
Standard definitions of negative feedback appear mostly to be based uponexperience derived from specialized and applied fields of engineering andcontrol systems. A more comprehensive theoretical view of the principles offeedback and control may also require more adequate nomenclature and lessrestrictive definitions.
Being very tentative, and harking back to an earlier thread in bit.sci-purposive-behavior,[CSG-L] and as an alternative to the term Control (which tends to suggestcontrol in isolation, by some agency of something else, I offer for discussionthe (made-up)term COMCONACTION, to connote e.g.:
COM -communication, common/together, selectively guided
CON -control, connect, con/with, consciousness (knowing together)
ACT and ACTION -actual event, selectively directed.
Comconaction is intended to suggest some independent action by some entity,as well as by groups of entities coordinated and guided by common purposes,i.e. activity of system-basedvector(s). The word is a little complicated, as is appropriate to what itexpresses.
Since a scientific term should be primarily denotative I suggest (again forpurposes of discussion): COMCONACTION is the integration of action in spaceand time, through negative feedback, of dynamic processes within systems at anylevel of organization or complexity, which enable relatively stable andpersisting output or behavior despite the effects of unpredictedcontingencies.
Other implications:
Since selectivity based upon criteria for error correction is inherent inthe irreducible primary conception and model for anything that can exist,structures required for evaluation and feedback must be present in any systemswhich continue to live and function.
While the values may be unspecified, the structural conditions exist, andthere may also be functional requirements for appropriate specific values.Indeed, many problems may ensue if the criteria utilized are inappropriate,e.g. at too low and restricted a level of the organization, rather thancriteria devised to reflect strategies which can benefit the whole. Indeedthe highest value may attach to the processes by which such strategies may bedeveloped -e.g. those required for a sustainable civilization on earth.
It is said by some that the highest values attach to the long termevolution of life and intelligence. However, in general, ideals which are tooabstract to be applied operationally do not provide useful criteria for theguidance of current decisions, i.e. they may not provide an ethical basis inpractice for existential judgements. The contention that such unapproachablevalues provide useful guidance has been historically fraught with disaster, sothat the onus is on the proponents to prove their case.
I would of course be interested in comments and any discussion. My wholepurpose is to try to suggest alternative approaches for consideration while thediscussion thread is still alive. However, I will be out of the country untilJune 24, and may miss Newsgroup postings, so I would appreciate email copies ofanything for which follow up or response from me is requested.
In any case, cheers and best wishes.
Bruce B.
Bruce Buchanan
*We are all in this together!*
Date: Sun, 4 Jun 1995 12:16:34 -0400
Subject: Re: Defining control
<[Bill Leach 950604.00:31]
>[Bruce Buchanan 950603:1540 EDT]
> A number of definitions have been cited. The problem with many of themis that they are of a technical nature and derivation, not at a level ofgeneral principle which carries well to other fields or has potentially largerimplications...
This is absolutely meaningless to me. Addition of real numbers is a veryprecisely defined term. It is technical in the extreme and combined with therest of the similarly strictly defined terms of mathematics provides us withthe most precise language that we have. Is it useful? I think that areasonable case can be made that it is useful.
This almost sounds like "Terms" must be ambiguous enough in meaning toallow one to make 'profound' statements without having to work veryhard."
The term "control" itself is "context sensitive" even here on CSG-Lwhere there really is an effort to maintain a precise meaning. For example myown recent use of the term without my having made a careful effort to explainthat my use was not quite in the "classic" (to PCT) meaning is likely the majorreason that Martin replied to my "control of a perception not currentlyperceived" discussion.
> The thesis I have in mind (not as original, but for emphasis anddiscussion) is that the most basic and primary level is the cybernetic model ofnegative feedback. Below this level of organization events do not exist for us(unknowable because unrepeated), there are no identifiable elements, nopossibility of patterned relationships to be conceived. Events and elementsbecome identifiable insofar as they participate and can be accessed repeatedlyand via feedback processes, which are essential to any structure ofrelationships involving time, hence any perceivable or knowable data. (I amprobably repeating thing Bill Powers has written somewhere.)
If Bill Powers wrote this somewhere then I certainly missed it. Indeed, ifI had read this in one of his works over a year ago I would still be asking himwhat he meant!
> ... More discriminating or selective perception reveals, however,that underlying and within primary experience are dynamic processes, includingperceptual processes, in constant renewal and interaction, both in the worldand within the organism. Persistence and renewal through time requires theaction of circular causal processes through which the enduring world and therepeatability and consistent identification of perceptions becomepossible.
That causal processes must exist to be able to develop a model of the worldI accept. That all processes must be circular causal processes for such tooccur I do not accept.
> ... (2) retroactive or circular causal loops which link elements andevents to repeat patterns in time, dampened depending upon Negative Feedback,but always with some delay (e.g. hysteresis) which is required and providesthe existential condition for possibilities for adaptive change. The relationwith time is crucial. Disturbance must actually occur, then be organized vianegative feedback, for random contingencies to be overcome, and for life toarise and endure.
Arg! Bruce, you are in an entirely different world from PCT here.
Patterns that repeat in time may not be retroactive causal loops withrespect to the observed pattern. The actual circular causality may not beobserved at all but rather only causal results (several times removed from the"ultimate" initiating cause).
In addition, generalizing "negative feedback" to exist in all processes isunreasonable for several reasons. In the first place while the limiting causedby system nonlinearities in oscillating system can be refereed to as negativefeedback, doing so without mentioning the nature of the system and the presenceof positive feedback in a significant portion of the operating cycle is highlymisleading.
Using the term "negative feedback" in reference to systems with an openloop gain of less than one is also not generally productive. Systems without acontinuous energy input are not well described in feedback terms.
Are you saying that "Delay" is hysteresis for negative feedback? Or thatthere is a hysteresis in loop transit time values? Your statement appearsmeaningless to me.
Disturbances to CEVs are resisted not organized by control systems.
The inorganic _exampleS_ that you give are arbitrary classifications byhumans. It is highly likely that these examples are interacting systems thatdefy analysis in isolation.
> So there are other views ( e.g. S-R)which see behavior as determined simply by What is perceived. Such views areemployed because they appear to the proponents be useful for special purposes,and may be validated by repetition within tightly constrained conditions....
The "special purpose" that you are referring to is, to the proponents, thesame one as for PCT --create a general theory of behavior.
> Terminology:
> Feedback
The precise mathematical definition for feedback in a closed loop controlsystem is THE ONLY proper definition for purposes related to PCT. That even weoften error in our verbal communications with respect to the meaning of theterm do not invalidate the correct meaning.
For some reason, there seems to be a general "movement" toward what I callrelaxing the definitions of terms. It seems that a significant number ofpeople want to be able to "say things" and "use terms" that will "make themsound impressive and knowledgeable" without also being subject to criticalscrutiny. That is they specifically do not want terms to have precise enoughdefinitions so that someone else could possibly make a valid challenge to whatthey say (ie: No one could claim that they might be in error).
While it might possibly be true that demands for strict use of terms mightbe a source for conflict between people, I personally view this situation asvastly superior to the situation where no one can hope to achieve even a modestunderstanding of the ideas of another because "it is pretty much OK for 'you'to mean whatever you want to mean by what you say"!
Even though it is true "that it's ALL perception" and that I, for example,can not know with certainty that I really understand what someone else isexpressing (at best), it is also equally true that it is the serious attemptsby one person to understand another that is the source of all of the knowledgethat is worth considering.
I personally now recognize that in practice, engineering use of the termcontrol is also sloppy. Even work in Control Theory tends to possiblyimproperly use the term. However, the meaning with respect to the fundamentalclosed loop control system is exacting.
Thus, I still believe that we are better off, to stick with this existingdefinition and point people to the source when needed rather than trying toinvent a new symbol.
In addition, I see your proposal as one that would create a significantlydiluted replacement.
The statement "which tends to suggest control in isolation, by some agencyof something else" IS A CORRECT DESCRIPTION OF THE PHENOMENON OF CONTROL fromthe viewpoint of an observer. "Some agency of a system is forcing some otherexternal 'thing' to maintain some constant condition, that is isolating this'thing' from the affects of other forces." Such is (correctly) the externalobservers view of what happens when a control system is functioning upon aCEV.
> Comconaction is intended to suggest some independent action by someentity, as well as by groups of entities coordinated and guided by commonpurposes, i.e. activity of system-basedvector(s). The word is a little complicated, as is appropriate to what itexpresses.
And by meeting this requirement set is useless for purposes of sayinganything "concrete", in my opinion.
> Since a scientific term should be primarily denotative I suggest(again for purposes of discussion): COMCONACTION is the integration of actionin space and time, through negative feedback, of dynamic processes withinsystems at any level of organization or complexity, which enable relativelystable and persisting output or behavior despite the effects of unpredictedcontingencies.
Control systems most explicitly DO NOT "enable relatively stable andpersisting output or behavior despite the effects of unpredictedcontingencies"!! It IS BEHAVIOR that is changed by control action to maintainthe desired perception!! Perception is what is constant in the presence of"unpredicted contingencies" or disturbances.
Arg!!
> Since selectivity based upon criteria for error correction is inherentin the irreducible primary conception and model for anything that can exist,structures required for evaluation and feedback must be present in any systemswhich continue to live and function.
And just where does this "axiom" come from?
Double Arg!!
> While the values may be unspecified, the structural conditions exist,and there may also be functional requirements for appropriate specificvalues.
for what?
> Indeed, many problems may ensue if the criteria utilized areinappropriate, e.g. at too low and restricted a level of the organization,rather than criteria devised to reflect strategies which can benefit thewhole.
Are you saying that a living control system that is "intended by design" tolive by foraging for berries in the forest that is equipped with only a "gilltype breathing system" is inappropriate? Also "whole" what?
> Indeed the highest value may attach to the processes by which suchstrategies may be developed -e.g. those required for a sustainable civilization on earth.
By whom and to what purpose? What is "sustainable civilization"? (givethree examples) :-)
> However, I will be out of the country until June 24, and may missNewsgroup postings, so I would appreciate email copies of anything for whichfollow up or response from me is requested.
I'm not sure that you would appreciate an email copy of this posting (justbe thankful that I too "ran out of time").
-bill
Date: Tue, 6 Jun 1995 02:44:18 -0600
Subject: control defs; abstractions
[From Bill Powers (950606.0000 MDT)]
Bruce Abbott (950605.1545 EST) --
Interesting quote from _Design of feedback control systems_, especiallythis:
> Such an open-loopcontrol system has the advantage of simplicity, but its performance is highlydependent upon the properties of the plant, which may vary with time.
The alleged "simplicity" depends to an extreme degree on the simplicity ofthe plant. If the plant has even a single time integration in it, the open-loopsystem has to be provided with computing capacity to calculate the inverse ofthe plant response to inputs, so that the reference signal can be passedthrough a computation that puts the inverse of the plant dynamics into thecontroller. For even modestly complex plants, calculation of the requiredinverses can soak up huge amounts of computing power and time. In general, afeedback model will accomplish the same result in a far simpler and faster wayrequiring very much less computation.
> [Bill P. will be happy to see this next one included:]
> 4. Increased speed of response and bandwidth. Feedback may be usedto increase the range of frequencies over which a system will respond and tomake it respond more desirably. A satellite booster rocket, for example, hasaerodynamics resembling those of a giant broomstick. It may, with feedback,behave with beauty and grace.
You bet I'm happy to see it. So how did this "feedback is too slow" shitget such a foothold in the literature?
---------------------------------------------------
Bill Leach (last few days or a week) --
I haven't been acknowledging your great posts much, but the reason is thatI get tired of saying yes, yes, yes.
One thing, though, that I must acknowledge: your comments on BruceBuchanan's latest offerings. It would be very easy to sit back and let youstick your neck out while I avoid the flak from saying the same things. But Iagree totally with your assessment of Buchanan's abstract wanderings. Withrespect to any kind of knowledge I'm interested in, or anything I'm interestedin doing, they are useless. Thanks for saying it so clearly.
Best to all, Bill P.