Truncated match.
PICList
Thread
'fuzzy logic'
1996\05\18@124711
by
Fernando Martin
1996\05\19@204603
by
fastfwd
Fernando Martin <.....PICLISTKILLspam@spam@MITVMA.MIT.EDU> wrote:
> Can anyone tell me where I might obtain a good (and cheap) fuzzy
> logic development tool for PIC16CXX.
Fernando:
Your local Microchip distributor sells two versions of Inform's
excellent "fuzzyTECHMP" tool. "fuzzyTECHMP Explorer" costs around
US$250; the fullfeatured "fuzzyTECHMP Edition" costs about three
times that. I have no idea what additional costs are imposed by
import duties, etc.
Andy
Andrew Warren  fastfwdKILLspamix.netcom.com
Fast Forward Engineering, Vista, California
http://www.geocities.com/SiliconValley/2499
1996\05\20@081504
by
rdmiller
On Sat, 18 May 1996, Fernando Martin wrote:
> Can anyone tell me where I might obtain a good (and cheap) fuzzy logic
> development tool for PIC16CXX.
Microchip's "fuzzyTECHMP Explorer" sells for about [US]$99 (that's
Part number DV005001, not the [US]$500 "Edition" DV005002). It comes
with a demo board to try your stuff out in hardware too.
The picture of the demo board shows an 18pin DIP socket for the PIC.
Uhoh... looks like it comes with a "Hardware Protection Key"
(read, "yet another dongle"). What a pain.
Rick Miller
'fuzzy logic'
1996\10\24@233032
by
TONY NIXON 54964
Has any body used the Fuzzy Tech development system?
The way I understand 'Fuzzy' programming is that the final code is
just a lot of IF ELSE THEN statements. Is this true?
Any help appreciated
Tony
Just when I thought I knew it all,
I learned that I didn't.
1996\10\25@014527
by
fastfwd
1996\10\25@025008
by
John Payson

> Has any body used the Fuzzy Tech development system?
>
> The way I understand 'Fuzzy' programming is that the final code is
> just a lot of IF ELSE THEN statements. Is this true?
>
> Any help appreciated
While I am not familiar with the particular fuzzy system you describe, a
typical fuzzy system loops through the following steps:
[1] All inputs are passed through "membership functions" which determine
to what extent they meet certain criteria. For example, in a thermo
stat controller, the "HOT" criterion might be 0% at 70% or below, 100%
at 90%, and intermediate values between. In some implementations, the
membership functions are computed at runtime (e.g. if thermister reads
X or beliw, return zero; if it reads Y or above, return 255; otherwise
return 255*(readingX)/(YX).
[2] The membership functions are run through a collection of "boolean"
equations. Rather than operating in the normal boolean sense, however,
and "AND" is evaluated as "minimum" and an "OR" is evaluated as "maximum".
[3] The outputs of the boolean functions are used to update the real oper
ation of the system (e.g. the amount of gas going to the furnace).
In terms of behavior, fuzzy systems generally convert abstract models of
a system into piecewiselinear approximations. For example, consider a
fuzzylogic controller for stopping a car at a precise point when the
mass of the car and effectiveness of the brakes is unknown. If the car
is moving very slowly (e.g. 1mph), the controller can pretty much just
let the car coast to the desired position and then apply the brakes. If
the car is travelling at, e.g. 5 mph it could wait until the last moment
at which it was _guaranteed_ to be able to stop the car and then hit the
brakes until the car was slowed to 1mph (at which point it would wait until
the car was in final position). If the car were traveling at 15mph, it could
wait until the last guaranteed stoping point, slow to 5, wait until the guar
anteed stopping point there, slow to 1, etc. While such a system would be
extremely jerky on a straightforward hardlogic system, using fuzzy logic
will make the system operate more smoothly (though still with the character
of performing the stop in multiple stages, with the reliability benefits that
would provide).
1996\10\25@221327
by
timetech
Yo!
Fuzzy logic doesn't use Boolean logic, at least not at the level of the
'fuzziness'; boolean logic is a restricted subset of fuzzy logic. Fuzzy
logic also does not produce piecewise linear approximation.
Trying to understand fuzzy logic by figuring out a model of how it works
based on other logical systems is largely futile. It leads to that
sinking feeling that there isn't really anything there. This has misled
quite a few otherwise sharp guys, including one of the real heros, Bob
Pease from National Semiconductor.
You have to try to understand the definition of fuzzy logic in its own
terms. This means that there are no obvious applications until you can
recast your problems into those terms. When you accomplish that, you
will see that there are whole classes of problems that are difficult or
intractable using standard methods that fall out elegantly using fuzzy
logic. In the end, it becomes just another handy tool in your kit of
tricks.
 Tom Rogers
1996\10\26@113450
by
Tom Messenger

At 10:11 PM 10/25/96 0400, you wrote:
<snip>
>Trying to understand fuzzy logic by figuring out a model of how it works
>based on other logical systems is largely futile. It leads to that
>sinking feeling that there isn't really anything there. This has misled
>quite a few otherwise sharp guys, including one of the real heros, Bob
>Pease from National Semiconductor.
>
>You have to try to understand the definition of fuzzy logic in its own
>terms. This means that there are no obvious applications until you can
>recast your problems into those terms. When you accomplish that, you
>will see that there are whole classes of problems that are difficult or
>intractable using standard methods that fall out elegantly using fuzzy
>logic. In the end, it becomes just another handy tool in your kit of
>tricks.
>
> Tom Rogers
<snip>
OK, I'll bite: what IS an application for fuzzy logic? I've seen articles
as you describe: apps that are described in terms of standard approaches. I
have yet to see someone suggest an application for fuzzy logic that either:
a) cannot be solved EASILY with conventional methods
or
b) isn't something grandiose such as hurricane prediction, etc.
If you know what you are talking about (and I hope you do), how about
presenting something to the PICLIST group on one application that can show
where fuzzy logic shines and other methods are NOTABLY lacking at least when
compared to fuzzy logic? And try to make it something that can be done at
the smaller end of the "complexity" scale... maybe that could be done with a
fuzzy controller on a "handheld" pcb (as opposed to solving chaos theory
problems on a mainframe...)
I have to conclude that there MIGHT be something to fuzzy logic (due to all
the marketing hype) but that it is such a strange approach thae even it's
adherents are almost but not quite completely unable to convey it's methods
clearly to the rest of the engineering community.
Tom Messenger
kristspam_OUTthegrid.net
1996\10\26@233242
by
Peter Grey

At 08:46 AM 26/10/96 0700, you wrote:
Tom,
To see where such logic is useful, a ride on elevators/lifts in Japan where
a lot of control is done by fuzzy logic will show the use of them. The same
applies to the trains. There is no jerky starts or stops but the actual car
moves faster between stops. I have tried this and can certainly vouch for
the performance. In regards to an aplication, there are many books out on
the subject and also many sources of information on the net. Just try a
search under one of the search engines.
If you are not successful here I will get hold of a few titles.
Good luck,
{Quote hidden}>At 10:11 PM 10/25/96 0400, you wrote:
><snip>
>>Trying to understand fuzzy logic by figuring out a model of how it works
>>based on other logical systems is largely futile. It leads to that
>>sinking feeling that there isn't really anything there. This has misled
>>quite a few otherwise sharp guys, including one of the real heros, Bob
>>Pease from National Semiconductor.
>>
>>You have to try to understand the definition of fuzzy logic in its own
>>terms. This means that there are no obvious applications until you can
>>recast your problems into those terms. When you accomplish that, you
>>will see that there are whole classes of problems that are difficult or
>>intractable using standard methods that fall out elegantly using fuzzy
>>logic. In the end, it becomes just another handy tool in your kit of
>>tricks.
>>
>> Tom Rogers
><snip>
>
>OK, I'll bite: what IS an application for fuzzy logic? I've seen articles
>as you describe: apps that are described in terms of standard approaches. I
>have yet to see someone suggest an application for fuzzy logic that either:
>
> a) cannot be solved EASILY with conventional methods
>or
> b) isn't something grandiose such as hurricane prediction, etc.
>
>If you know what you are talking about (and I hope you do), how about
>presenting something to the PICLIST group on one application that can show
>where fuzzy logic shines and other methods are NOTABLY lacking at least when
>compared to fuzzy logic? And try to make it something that can be done at
>the smaller end of the "complexity" scale... maybe that could be done with a
>fuzzy controller on a "handheld" pcb (as opposed to solving chaos theory
>problems on a mainframe...)
>
>I have to conclude that there MIGHT be something to fuzzy logic (due to all
>the marketing hype) but that it is such a strange approach thae even it's
>adherents are almost but not quite completely unable to convey it's methods
>clearly to the rest of the engineering community.
>
>
>Tom Messenger
>
@spam@kristKILLspamthegrid.net
>
>
Peter Grey
Australia
1996\10\27@040739
by
Shel Michaels
In a message dated 961026 12:06:46 EDT, Tom Messenger writes:
<< how about
presenting something to the PICLIST group on one application that can show
where fuzzy logic shines and other methods are NOTABLY lacking at least when
compared to fuzzy logic? And try to make it something that can be done at
the smaller end of the "complexity" scale.
>>
Hear! Hear!!
...Shel Michaels
KILLspamsbmichaelsKILLspamaol.com
1996\10\27@203524
by
Steve Hardy

> From: Tom Messenger <RemoveMEkristTakeThisOuTTHEGRID.NET>
> [cut]
> OK, I'll bite: what IS an application for fuzzy logic? I've seen articles
> as you describe: apps that are described in terms of standard approaches. I
> have yet to see someone suggest an application for fuzzy logic that either:
>
> a) cannot be solved EASILY with conventional methods
> or
> b) isn't something grandiose such as hurricane prediction, etc.
>
> [cut]
> I have to conclude that there MIGHT be something to fuzzy logic (due to all
> the marketing hype) but that it is such a strange approach thae even it's
> adherents are almost but not quite completely unable to convey it's methods
^^ huh?
> clearly to the rest of the engineering community.
I too am a bit skeptical. The latest Circuit Cellar Ink magazine had a
few articles. One example was for a crane controller for unloading
ships at dock. The aim was to stop the crane at the correct spot,
without load swing. The author claimed that the conventional approach
required solving a 5thorder differential equation, whereas the fuzzy
logic approach used a bunch of 'rules of thumb' (rule of thumbs if I
was the operator).
One of these days I'm going to make a birobalancer. This is a small pad
with servo control in the X and Y directions which balances a biro on
its ball point. Once I work out a suitable feedback sensor, I will
implement the controller code using conventional and FL approaches and
report back which is better/easier.
Certainly, a ROT system would be easier to understand than even a
single ODE. ODE solvers would be very difficult to implement in a
microcontroller with no floatingpoint, whereas fuzzy logic seems to
require only multiply and add (fixed point, if any).
Both solutions would need tuning, but I think the tuning would be
easier for fuzzy logic, and probably more robust. DEs and control
loops are notorious for being sensitive to parameter variations,
initial conditions and deviations from assumptions of linearity.
After chanting the universal mantra, Om, for a few hours, my head
cleared sufficiently for it to be able to consider the benefits of
fuzzy logic. I think some of the bad press surrounding FL is because
of the name, which sort of looks like 'fuzzy head' or 'drunken stupor'
or 'programmer who doesn't know what he's doing'. Nevertheless, like
'object oriented' there's no need to dismiss a way of thinking just
because it's 'soup du jour'.
Regards,
SJH
Ailartsua, Arrebnac
1996\10\28@115242
by
John Piccirillo

With the usual disclaimer that I'm not a fuzzy logic expert, or even
practitioner, fuzzy logic allows the solution of certain mathematical
problems with using a math model. Here are two simple examples that are
frequently given (NOT the room temperature example).
1) Balancing an inverted pendulum. This problem can be solved using a PID
controller, but it's fuzzy logic solution is much simpler. So much so that,
in simulation, fuzzy logic has been used to solve double (and I think
triple) inverted pendulums. One doesn't have to know anything about
rotational inertia, etc. Essentially two variables are needed, position
angle from vertical and rate of change of the angle. These are binned into
categories such as small, medium, large, positive fast, etc  a bunch of
linguistic variables. The measured (or simulated) variables are put into
these categories via membership functions (which are not exclusive, ie it
can be mostly warm but also slightly cool. If then rules are applied to
turn the fuzzified inputs into outputs, and the fuzzified outputs are
combined and defuzzified, telling the controller what to do. Notice that
there is no explicit math model.
2) Backing a tractor trailor into a parking slot. Truck drivers do this
without explicitly solving algorithms. There procedures were taken from
interviews and turned into If Then rules. The above procedure was used to
produce a fuzzy logic solution, which works well in real time.
Two beginning, nontechnical, books that explain the concepts are Fuzzy
Logic, and Fuzzy Thinking.
{Quote hidden}>
>Date: Sat, 26 Oct 1996 08:46:55 0700
>From: Tom Messenger <
spamBeGonekristspamBeGoneTHEGRID.NET>
>Subject: Re: Fuzzy Logic
>
>OK, I'll bite: what IS an application for fuzzy logic? I've seen articles
>as you describe: apps that are described in terms of standard approaches. I
>have yet to see someone suggest an application for fuzzy logic that either:
>
> a) cannot be solved EASILY with conventional methods
>or
> b) isn't something grandiose such as hurricane prediction, etc.
John
1996\10\28@132433
by
Walter Banks

Tom Messenger wrote:
>
> OK, I'll bite: what IS an application for fuzzy logic? I've seen articles
> as you describe: apps that are described in terms of standard approaches. I
> have yet to see someone suggest an application for fuzzy logic that either:
>
> a) cannot be solved EASILY with conventional methods
> or
> b) isn't something grandiose such as hurricane prediction, etc.
>
It took us two years before we discoved what fuzzy logic does well. We fell
into the trap of trying to implement things that worked well in a
PID control system and then tried to reimplerment the same problem
in Fuzzy logic.
We have found that fuzzy logic works well in non linear systems. I can give
you a simple example of this. Take an home environment control system where
the home as both airconditioning and furnace as well as external daily sun or
night. The home requirements are also have a daily and weekly cycle. The
control requirement is to control temperature and humidity with some attention
to energy requirements.
The fuzzy logic system approaches the problem as several separate independent
control problems whose results are weighted. Control (furnace and
airconditioner)
is based on simple decisions on the weighted fuzzy logic control results. The
non linear nature of the problem,(outside cloud or sun , changes in termal
load,
Effect of outside temperature on inside humidity) all make for complex
conventional solutions.
Such a system can be easily implemented with fuzzy logic in a 16C74 whereas
the convetional solution would be considered a challange in the same part.
Walter Banks
http://www.bytecraft.com
1996\10\28@142035
by
Chuck McManis

> The non linear nature of the problem,(outside cloud or sun ,
> changes in termal[sic] load, Effect of outside temperature on
> inside humidity) all make for complex conventional solutions.
This is a misuse of the term 'nonlinear'. A nonlinear weather system would
go from extremely hot, to snowing, instantaneously. That would constitute a
nonlinear change in temperature. Complex certainly, but also just as certainly
a linear system.
Someone else here hit it right on the head, "Fuzzy Logic" is simply a design
methodology like "object oriented" is simply a programming methodology. One
methodology may make specifying the solution easier (try specifying a C
compiler in assembly language, its possible but painful) but both methods
result in code running on a microprocessor. Which is "best" can be quantified
in terms of:
1) Which system would allow the engineer to specify the solution
to the current problem in less time.
2) Which system would be more likely to produce a "correct" or working
result the first time?
3) Which system would me the resource allocations available, which
would beat the allocations?
4) Which system would be easier to modify when the engineer who
built it left?
There is no mystery here. Another thing that "fuzzy logic" proponents gloss over
is the difficulty of coming up with good membership functions. This is just like
the "object oriented" proponents glossing over the fact that getting the right
encapsulation model is quite hard. Assembly language hackers quote speed
and gloss over how difficult it is to train a new engineer to replace them, or
to
change platforms. Bits is bits.
Chuck
1996\10\28@154126
by
Clyde SmithStubbs

Chuck McManis <TakeThisOuTcmcmanisEraseMEspam_OUTFREEGATE.NET> wrote:
> This is a misuse of the term 'nonlinear'. A nonlinear weather system would
> go from extremely hot, to snowing, instantaneously. That would constitute a
> nonlinear change in temperature. Complex certainly, but also just as
certainly
> a linear system.
I think you're confusing "nonlinear" with "discontinuous". A nonlinear
function
is one which graphically is not a straight line. A discontinuous function is
one where there "steps" or "breaks" in the transform, e.g. the tangent function
is nonlinear for any value, but discontinuous only at theta=pi. It also happens
to be undefined for that value, but that's another issue.
A discontinuous function must be, by definition, nonlinear, but the reverse
is not true.
However, I'm not sure that Walter's description of the environmental control
system as a "nonlinear" system is particularly useful. There are very few
problems in the realworld that are linear! It may be that the conflicting
goals inherent in that system are the factor that makes it suitable for
fuzzy logic (this would fit with many of the other mentioned applications,
too).
Clyde

Clyde SmithStubbs  HITECH Software,  Voice: +61 7 3354 2411
RemoveMEclydeTakeThisOuThitech.com.au  P.O. Box 103, Alderley,  Fax: +61 7 3354 2422
http://www.hitech.com.au  QLD, 4051, AUSTRALIA. 

For info on the World's best C cross compilers for embedded systems, point
your WWW browser at http://www.hitech.com.au, or email infoEraseME.....hitech.com.au
1996\10\28@164242
by
Walter Banks

Chuck McManis wrote:
>
> > The non linear nature of the problem,(outside cloud or sun ,
> > changes in thermal load, Effect of outside temperature on
> > inside humidity) all make for complex conventional solutions.
>
> This is a misuse of the term 'nonlinear'. A nonlinear weather system would
> go from extremely hot, to snowing, instantaneously. No that is a system with a
step function.
> That would constitute a nonlinear change in temperature. Complex certainly,
> but also just as certainly a linear system.
>
> Someone else here hit it right on the head, "Fuzzy Logic" is simply a design
> methodology like "object oriented" is simply a programming methodology.
Fuzzy logic is actually two things.
1) A design methodology that offers a different way of implementing a
solution to a problem.
2) Fuzzy logic has a mathematical basis that is not found in conventional
systems. When used in the context of a control system the convention
implementation assumes that the parameters of a system are constant and
the response is linear. Fuzzy logic allows developers to describe a
solution
that doesn't need that assumption. Case in point is a speed control system
where the mass of the object can be independent of the control system. A
practical example is the control system of an airplane where the flight
mass and angular momentum varies over duration of a flight.
The math behind fuzzy logic is really the manipulation and use of
linguistic
variables. It is this additional data type that differentiates Fuzzy
from Crisp implementations.
> One methodology may make specifying the solution easier (try specifying a C
> compiler in assembly language, its possible but painful) but both methods
> result in code running on a microprocessor.
Don't confuse implementation pain with implementation approach. Fuzzy offers
both. It is implementation approach that is the important one. Much of the
work that I seen in Japan implements a fuzzy control solution in assembly code.
> Which is "best" can be quantified in terms of:
I can answer this assuming a process control system.
> 1) Which system would allow the engineer to specify the solution
> to the current problem in less time.
Except for trivial problems the fuzzy solution is more intuitive.
> 2) Which system would be more likely to produce a "correct" or working
> result the first time?
My experience is the fuzzy based system. I have more experience with
conventional
PID systems. Linguistic variables allows me to focus on the solution rather
that implementation details.
> 3) Which system would me the resource allocations available, which
> would beat the allocations?Both systems.
> 4) Which system would be easier to modify when the engineer who
> built it left?
In a control system the fuzzy implemented system. I have built control systems
both ways, by far the fuzzy implementation is easier to maintain. The biggest
advantage of maintaining someone else's code is the fuzzy implementation is less
likely to break with changes.
>
> There is no mystery here. Another thing that "fuzzy logic" proponents gloss
over
> is the difficulty of coming up with good membership functions.
Membership functions are easier to get right than feedback constants in a PID
control system. Membership functions have not been a problem in any of the
applications that I have worked on.
Walter Banks
http://www.bytecraft.com
1996\10\28@175057
by
Chuck McManis

Walter wrote:
> 2) Fuzzy logic has a mathematical basis that is not found in conventional
> systems.
If "conventional systems" means systems composed entirely of algebraic notation
I agree with you, however if you read a "fuzzy" logic text and a group theory
mathematical text you will quickly be able to translate between the two. The
other area of software that this stuff is pretty prevalent in is the logical
constraint
programming.
Note that I hold no bias either for or against fuzzy logic. I took the time a
while
back to become intimately familiar with the process, the terminology, and the
roots of the methodology. It can be used to express a set of feedback control
parameters more intuitively than perhaps some other methods, once
reduced to practice it cannot produce a "better" result than a tradiional
approach from someone skilled in the art. It can also produce wildly worse
solutions if misapplied (as does anything, no disrespect intended here.)
Chuck
1996\10\28@181432
by
timetech
Steve Hardy wrote:
(snip)
> After chanting the universal mantra, Om, for a few hours, my head
> cleared sufficiently for it to be able to consider the benefits of
> fuzzy logic. I think some of the bad press surrounding FL is because
> of the name, which sort of looks like 'fuzzy head' or 'drunken stupor'
> or 'programmer who doesn't know what he's doing'. Nevertheless, like
> 'object oriented' there's no need to dismiss a way of thinking just
> because it's 'soup du jour'.
>
> Regards,
You'll love the new term for fuzzy logic that is starting to appear:
"Zadehean logic", after Lofti Zadeh, one of the original fuzzy
theorists.
 Tom Rogers
1996\10\29@093922
by
Walter Banks

Chuck McManis wrote:
>
> Walter wrote:
> > 2) Fuzzy logic has a mathematical basis that is not found in conventional
> > systems.
>
> If "conventional systems" means systems composed entirely of algebraic
notation
> I agree with you, however if you read a "fuzzy" logic text and a group theory
> mathematical text you will quickly be able to translate between the two.
Fuzzy logic doesn't compete with conventional systems (Non Fuzzy). In practice
fuzzy logic is the manipulation of lingustic variables. Lingustic variables
are not normally found in comventional systems.
> Note that I hold no bias either for or against fuzzy logic. I took the time
> a while back to become intimately familiar with the process, the
> terminology, and the roots of the methodology. It can be used to express
> a set of feedback control parameters more intuitively than perhaps some
> other methods, once reduced to practice it cannot produce a "better" result
> than a tradiional approach from someone skilled in the art.
All depends what the definition of "better" is. Appropriate use of lingustic
variables can reduce the complexity of a solution. Lingustic variable
manipulation is often less computation intensive than calculations with
their crisp conterparts. A parallel is the relationship between floating
point and fized point math.
As a Japanese friend of mine said over diner a few months ago in Tokyo, "Fuzzy
logic is now another tool in an extensive tool box". He is a person who has
implemented many applications some using both linguistic variables and
conventional variables and logic. For me it has allowed me to easily solve
some problems that were extremely difficult to find solutions with other
techniques available to me.
I suppost it finally boils down to using linguistic variables in
applications or not. I use them where appropriate.
Walter Banks
1996\10\30@002332
by
Dave Williams

The biggest problem I have with fuzzy logic is after the design, there is no
"figure of merit". In conventional control theory, you have stability criteria
like phase margin or gain margin. In fuzzy, it just seems to work, but it does
not tell you how well it works i.e. how stable the compensated system is. Yea,
I know the poles and zeros move around but at least you have an estimate if you
are on the ragged edge of stability with conventional control theory.
Like it or not, I think it is very difficult to design a control system without
some using concepts from conventional theory take bandwidth for example. If
you have a system with a BW of 1000 hz. (from measurements, calculations what
have you) you better look for feedback transducers that have a response greater
that 1000 hz. I am at a loss to understand how fuzzy would help a designer in
choosing transducers for example. Or understanding frequency response in input
signal conditioning. I have been told that some fuzzy designers after
completing the design run simulations to determine BW, etc. But that seem
backwards to me. Maybe the fuzzy experts out there can answer these questions
or point me in the right direction.
Dave Williams
Lakewood, CO.
1996\10\30@100413
by
timetech

Dave Williams wrote:
>
> The biggest problem I have with fuzzy logic is after the design, > there is
no "figure of merit". In conventional control theory, you > have stability
criteria like phase margin or gain margin. In fuzzy, > it just seems to work,
but it does not tell you how well it works i.e. > how stable the compensated
system is...
Well, the figures of merit you mention are applicable to the linear math
model; they don't apply to the fuzzy logic model. Of course, there are
practical figures of merit for any system, like does it oscillate or
overshoot or crash into the dock or whatever.
In most PID control loops the theoretical figures of merit are never
calculated. Instead, the technician from the instrument shop known as
'Magic Fingers' is called in to tweak the controller constants. Many
modern PID controllers have 'Magic Fingers' built in; some of them are
now using fuzzy logic to adjust the control parameters.
With fuzzy logic it is relatively simple to identify the stability
parameters of interest and build them into the system, which then
becomes psuedo adaptive. If you provide a mechanism for adjusting the
rules than the system is really adaptive. There are several techniques
for doing this, including using a neural net in conjunction with the
fuzzy logic model.
The psuedo adaptive model is one place that fuzzy logic shines over
standard PID implementations. I've looked at some of the temperature
controller applications (refrigerators and air conditioners, mostly) and
the big improvement is in adaptive tuning to a changing load, to put it
into standard speak. I programmed a bunch of systems using PID
algorithms that were embellished to handle load variations, and I think
the fuzzy solution is probably simpler and more resilient.
Each example of the PID code was specific to a certain system (we're not
just talking batching control here; the systems were fairly complex).
Although the solution space adequately covered the system input range in
each case, I wouldn't expect the systems to be interchangable. In fact,
I would expect stability problems if I did that.
The fuzzy solution, on the other hand, already has the entire solution
space covered, just by introducing the relevent rules. The rules might
be mistuned, but there will be some response, even if less than optimal.
A PID loop with a bunch of ifthenelse style expert additions usually
has edges in the solution space beyond which there is ill defined
output.
Of course, 99.99% of those temperature loops out there are still under P
or PI or PID control, and work pretty good...
 Tom Rogers
1996\10\31@165615
by
Martin McCormick

I thought that the concept of fuzzy logic was simply that the rules
for solving a problem change based on inputs from the environment. As an
example, we are in a time of year when some heating is required to make the
house comfortable. The problem is that there is a great variation from day
to day as to current weather conditions so that if one uses a timerbased or
setback thermostat to turn up the heat in the morning, one runs the risk of
having heat on when the outside temperature is not cold. A fuzzy thermostat
controller would have a setback schedule in it just like the nonfuzzy kind,
but it would also have a thermistor outside to read the current temperature
and would not apply any heat if that were above a certain set point. The
fuzzy controller might get input from pin switches in window frames to see
if the residents of the house had opened a window or two during the night
since it would be stupid to turn on any heat if that were the case.
A fuzzy lawn sprinkler controller would use a timer, but only actually
turn on the water if it was not raining already, an algorithm which appears
to escape the protoplasmbased controllers in the heads of the grounds crew
on our campus.
Martin McCormick WB5AGZ Stillwater, OK 36.7N97.4W
OSU Center for Computing and Information Services Data Communications Group
1996\10\31@171839
by
Walter Banks

Martin McCormick wrote:
>
> I thought that the concept of fuzzy logic was simply that the rules
> for solving a problem change based on inputs from the environment. As an
> example, we are in a time of year when some heating is required to make the
> house comfortable. The problem is that there is a great variation from day
> to day as to current weather conditions so that if one uses a timerbased or
> setback thermostat to turn up the heat in the morning, one runs the risk of
> having heat on when the outside temperature is not cold. A fuzzy thermostat
> controller would have a setback schedule in it just like the nonfuzzy kind,
> but it would also have a thermistor outside to read the current temperature
> and would not apply any heat if that were above a certain set point. The
> fuzzy controller might get input from pin switches in window frames to see
> if the residents of the house had opened a window or two during the night
> since it would be stupid to turn on any heat if that were the case.
Earlier today I posted a fuzzy code fragment for a home control system
similar to what you are suggesting. It has linguistic variables for
the time of day. (morning, day, evening, night) as well as for inside
and outside temperatures. It also has some "green" rules that will not
turn on the furnace if the outside temperature is hotter than the inside
the oposite for airconditioning. In fuzzy's style all the rules are
evaluated and a weighted result is produced.
Walter Banks
1996\10\31@221408
by
Martin McCormick
I see I have basically repeated what Walter Banks said in an earlier
message. I missed that one and am sorry for the duplication.
In message <EraseME3275236B.6ACAbytecraft.com>, Walter Banks writes:
>
>It took us two years before we discoved what fuzzy logic does well. We fell
>into the trap of trying to implement things that worked well in a
>PID control system and then tried to reimplerment the same problem
>in Fuzzy logic.
>
>We have found that fuzzy logic works well in non linear systems. I can give
>you a simple example of this.
Martin McCormick WB5AGZ Stillwater, OK 36.7N97.4W
OSU Center for Computing and Information Services Data Communications Group
'fuzzy logic'
1996\11\02@101855
by
Gerhard Fiedler

At 00:20 30/10/96 EST, Dave Williams wrote:
>Like it or not, I think it is very difficult to design a control system without
>some using concepts from conventional theory take bandwidth for example. If
>you have a system with a BW of 1000 hz. (from measurements, calculations what
>have you) you better look for feedback transducers that have a response greater
>that 1000 hz. I am at a loss to understand how fuzzy would help a designer in
>choosing transducers for example. Or understanding frequency response in input
>signal conditioning. I have been told that some fuzzy designers after
>completing the design run simulations to determine BW, etc. But that seem
>backwards to me. Maybe the fuzzy experts out there can answer these questions
>or point me in the right direction.
I haven't worked with fuzzy yet (but with "conventional" methods), but it
seems that a fuzzy design usually is based on a description of a _working_
solution in what they call lexical terms, i.e. like the example of the truck
moving backwards: you do not have to analyse the problem mathematically,
because you use an approximation to the (already proven to work) rules of a
(some) truck driver(s).
It seems obvious that you need some understanding (or a good understanding)
of the process involved, but not necessarily on a mathematically analytical
base. A good truck driver will bring that monster backwards in if possible
at all, and he wouldn't know to tell you the mathematics to be able to solve
it "classically". But he has a quite profund understanding of the process,
on another "plane", and maybe he can give you a set of those "lexical" rules.
It seems that fuzzy helps to use just these forms of understanding of a
process. With conventional methods it isn't easy at all to include the
intuitive knowledge of the people who often have worked successfully
controlling a process in a "manual" way, and you may end up not taking much
advantage of this knowledge. This I think is one of the major drawbacks to a
classical mathematical approach, which fuzzy may help to avoid.
'Fuzzy Logic'
1997\10\02@213302
by
Philip Starbuck
I have a project that I believe lends itself to the use of Fuzzy Logic
implemented on a PIC. As I recall MICROCHIP promoted their Fuzzy TchMp
Pic16/17 Explorer as a low priced introduction into the area of Fuzzy
Logic. However, when I called the local MICROCHIP office I received the
all to typical of late, "DUH?, I dunno do we sell that?, Try a distributor"
answer. Good ol DIGIKEY took a week to quote a price equlivent to a low
price pentium system for the "introductory" kit.
Is this the only way, or the most practical way to investigate the use of
Fuzzy Logic or there more cost effictive alternatives?
cheers,
Phil
Philip Starbuck
(909) 7927917
"There are three principal ways to lose money. Wine, women, and engineers.
While the first two are more plesent the third is by far the more certain."
 Baron Rothschild
ca. 1860
'Fuzzy Logic'
1997\11\08@095516
by
Tom Handley
The recent issue of Circuit Cellar Ink (#88, Nov 97) is covering Fuzzy Logic
again. Of interest to many in this group is an article by Mr Byte Craft,
Walter Banks, discussing Fuzzy concepts using C. It would be nice to hear from
folks implementing Fuzzy Logic with the PIC family.
 Tom
1997\11\08@120831
by
Walter Banks
> The recent issue of Circuit Cellar Ink (#88, Nov 97) is covering Fuzzy
Logic
> again. Of interest to many in this group is an article by Mr Byte Craft,
> Walter Banks, discussing Fuzzy concepts using C. It would be nice to hear
from
> folks implementing Fuzzy Logic with the PIC family.
There is actually quite a lot of fuzzy logic being done with PIC's.
Microchip
has been distributing Inform's fuzzy logic development tools for analyzing
and
creating fuzzy logic based applications.
We have Fuzzc which is a general purpose C preprocessor that adds
Linguistic variables and their manipulation to any C program. FuzzC passes
normal C straight through and translates the Linguistic variables and
operations
into normal C.
Most of the applications that I see being done on the PIC's
are nonlinear control systems.
Walter Banks
http://www.bytecraft.com
1997\11\08@172102
by
Walter Banks
> The recent issue of Circuit Cellar Ink (#88, Nov 97) is covering Fuzzy
Logic
The significance of the Circuit Cellar article is the math behind making
close
comparisons with crisp numbers. It shows a number of ways to mix crisp
and lingustic variables in the same expression.
Fuzzy logic may often reduce the amount of code and execution cycles
needed to solve a problem some problems. Good canidates for this
kind of savings is the type of problems that start with complex data or
data with a large dynamic range producing a simple result. This is one
of the reasons that fuzzy logic is very effective in nonlinear control
systems and pattern matching.
Walter Banks
1997\11\10@074322
by
Tom Handley

Walter, I've spent a lot of years designing energy management systems
for commercial buildings before there were Fuzzy tools. It's a natural
application for Fuzzy logic with the classic thermostat problem as well
as humidity and loadshedding issues. I'd love to get back into it given the
hardware and software available today. At the time we used the 6809 and a
Forthbased multitasking kernel on the STD bus.
 Tom
At 12:05 PM 11/8/97 0500, you wrote:
{Quote hidden}>> The recent issue of Circuit Cellar Ink (#88, Nov 97) is covering Fuzzy
>Logic
>> again. Of interest to many in this group is an article by Mr Byte Craft,
>> Walter Banks, discussing Fuzzy concepts using C. It would be nice to hear
>from
>> folks implementing Fuzzy Logic with the PIC family.
>
>There is actually quite a lot of fuzzy logic being done with PIC's.
>Microchip
>has been distributing Inform's fuzzy logic development tools for analyzing
>and
>creating fuzzy logic based applications.
>
>We have Fuzzc which is a general purpose C preprocessor that adds
>Linguistic variables and their manipulation to any C program. FuzzC passes
>normal C straight through and translates the Linguistic variables and
>operations
>into normal C.
>
>Most of the applications that I see being done on the PIC's
>are nonlinear control systems.
>
>Walter Banks
>
http://www.bytecraft.com
>
>
'Fuzzy Logic'
1998\09\26@144228
by
shadedemon
Peter L. Peres wrote:
> The processor implements a whole set of process equations, as opposed to
> one for PID et al. At any given moment only one equation functions, but
> the equation can be changed depending on the exact system parameters at
> that point. A state machine in the processor keeps track of the
Hmm I've always seen it as the parameters picking a subset
of the equations to operate, not just one. Depends on what
you're doing and how you specify it..
The BEST article I've run across was in the MicroComputer
Journalbefore the whirlwind publications lack of
publication. November/December 1994, p 17. The graphs are
screwed up so can confuse if you don't recognize it, but the
text is clear.
Alan
1998\09\26@151505
by
Peter L. Peres
On Sat, 26 Sep 1998, Alan King wrote:
> Peter L. Peres wrote:
> > The processor implements a whole set of process equations, as opposed to
> > one for PID et al. At any given moment only one equation functions, but
> > the equation can be changed depending on the exact system parameters at
> > that point. A state machine in the processor keeps track of the
>
> Hmm I've always seen it as the parameters picking a subset
> of the equations to operate, not just one. Depends on what
> you're doing and how you specify it..
The parameters eventually pick just exactly one equation eventually, but
which one of the possibles for that Ypoint, depends on the present state
and on the previous (one or more) state(s). This is not so obviously
stated in books imho, and it is essential to understand the concept imho.
Peter
More... (looser matching)
 Last day of these posts
 In 1998
, 1999 only
 Today
 New search...