Fact is that FL apparently works in areas where PID struggles, with non-linearities, noise, etc. I've never used it in anger, but I've had my fair share of issues with PID instability in the past so would probably give it a go in the future...

 

The biggest gripe was the way that FL was touted as being the solution to every problem, and that all who did not immediately embrace FL completely were, at best, unqualified.

 

If it does the job as intended and requires quite a bit less programming for the case at hand then I am all for it !

 

Really, PID unning with "Big", "some", and "a little bit", instead of actual exact numbers , could probably work very well as a feedback controller.

So some kind of bastard hybrid between the two?

The biggest gripe was the way that FL was touted as being the solution to every problem, and that all who did not immediately embrace FL completely were, at best, unqualified.

There's always some hype whenever a new technique is developed.
Doesn't mean that technique has no merit.

 

I've had my fair share of issues with PID instability in the past

PID tuning is a bit of a Black Art and Foxboro even developed an Auto Tune for PID in their software and field interface hardware. But the only time I ever had a complete failure was when some idiot grabbed some control valves from surplus and used them on a project. 2" line needs a 2" control valve, got one, check. But never did the flow calculations or trimmed the valve internals for the correct flow parameters. Not the PID control failure but complete lack of engineering understanding of flow control and failure to do the correct engineering required to control it. Another instance was Ph control and once again not the controls but the basic instability of Ph sensors and tendency of probes to become fouled by the goop they were inserted into. Temperature control failure in a boiler due to temp sensor failure due to soot or ash caking, etc. Sometimes it's not the controls themselves at fault but the inherent instability of the system they are trying to control.

 

Really, PID unning with "Big", "some", and "a little bit", instead of actual exact numbers , could probably work very well as a feedback controller.

So some kind of bastard hybrid between the two?

I don't see how that would work. PID is based on a computed mix of three continuous and mathematically/physically related parameters related to the difference between an actual outcome and a desired outcome. FL is more a classification of input(s) into one or more sets that computes a likely response.

 

Certainly NOT EVERY INSTANCE would work, and probably none as good as an optimized and calibrated PID system.
Even an ON/OFF system canwork if the control response time is adequately faster than the sensor response time.

 

There are cases where it is equally as good or even better in some instances, all documented.
I am not ready to dismiss it, if it does the job.
(We do it every time we drive a car !)

 

Certainly NOT EVERY INSTANCE would work,

And in WHAT INSTANCE would it not work?
Or is that just your OPINION?

probably none as good as an optimized and calibrated PID system.

With the caveat that the system has small non-linearities and little dead-zone/backlash.

 

One standard product( a testing machine) that reqired an exqct pressure be held, in a large volumre air tank, which supplied the compressed air to drive a crash-sled at a very specific velocity. The maximum accelleration had to be limited to avoid "jostling" the crash dummies, as well. What makes setting air pressure tricky is that as the pressure is adjusted, the temperature tends to change in response, which tends to change the pressure. So the control scheme had to include the temperature at the time the pressure was adjusted. so the process got complicated.

 

What makes setting air pressure tricky is that as the pressure is adjusted, the temperature tends to change in response, which tends to change the pressure. So the control scheme had to include the temperature at the time the pressure was adjusted. so the process got complicated.

Sounds like that would have been a good candidate for Fuzzy Logic control.

 

My problem with the "Fuzzy" is that I have never seen an algorithm that defines implementing it- it's all talk.

 

Hello,

I found the attached pdf about fuzzy control design.

Bertus

 

From that PDF it seems that the whole concept is based on using "greater than" and "less than" types of information, rather that exact numerical data. That is hardly a new concept, although it may be unfamiliar to some folks.
Adjusting system loop gain and response based on error magnitude categories can certainly be simpler than a constant gain PID servo system.

 

Adjusting system loop gain and response based on error magnitude categories can certainly be simpler than a constant gain PID servo system.

I don't have a dog in this fight: I know nothing of Fuzzy Logic.

But, in my PID (both analog and digital) designs, the error is always (vanishingly) small. In other words, if there is more than a very small (AC) error, the loop is not tracking correctly.

How does one account for this with IF/THEN statements?

 

Here is a really good video that explains it.

 

it seems that the whole concept is based on using "greater than" and "less than" types of information, rather that exact numerical data.

But the "greater than" or "less than" can go down to the resolution of the digital signal data if desired, which is the same "exact" accuracy limit of a digitized PID loop.

PID is an analog control system that has served us well for over a hundred years, but suffers from problems with integrator windup, or with non-linear systems and/or loop hysteresis.
With the advent of digital processors, it makes sense to seriously consider a control loop optimized for digital which is not an analog process emulated in digital, and that takes advantage of digital processing doing things that an analog processor can't readily handle, .

PID works well with analog processing in a linear control system, since op amps can readily and easily do rapid integration and differentiation.
Computers not so much.

 

Anyone here knows if drones use PID or Fuzzy Logic in their programming?

 

ORIGINALLY, ALL of the flight control computing was done INSIDE the pilot's head. Then instruments got smaarter and actual srvo-systems arrived. Proportional control with derivative assistance was a load reduction assistant of great value. Compensating for errors from different sources was done by the pilot. And possibly the navigator helping.
As for fuzzy logic in drones, unfortunately flight control is terribly unforgiving of errors in most instances.

 

As for fuzzy logic in drones, unfortunately flight control is terribly unforgiving of errors in most instances.

So have you some known reason for implying that FL will have more errors in controlling a drone's flight as compared to PID?
Or are you just pontificating as usual?

Anyone here knows if drones use PID or Fuzzy Logic in their programming?

According to Goggle AI:
Yes, drone software utilizes fuzzy logic to make robust, real-time navigational and control decisions. It is particularly valuable for handling dynamic, unpredictable environments—such as turbulent wind, sudden obstacles, or landing on a moving platform—without requiring highly complex mathematical models