Tutorial A layman's analogy guide to PIDs

Excellent post @noorbeast - It's certainly lifted the cloak from the PID for me. I never understood the relationships and just blindly entered numbers until the arm moved in a way that looked OK. Love the analogy! I always revert to cars when I explain computer related issues!

My favourite is when the powers that be don't understand why a system need to come down for maintenance... I reply with "Would you keep your car engine running while the mechanic worked on the pistons??"

Thanks again @noorbeast

Actually... I was just looking at my settings and the P and D values are quite close to each other. (as per @Avenga76 'settings in his thread)

I assume the actual numbers don't really matter that much so long as the desired graph is close to the target without a huge overshoot or over-dampen?

@noorbeast - Would you be able to maybe make a short video where you tune a JRK from scratch to highlight the relationships? I know it's asking a lot. What can I say.. I'm a visual learner!

Thats right @Archie , everyones numbers will be different. And with different controllers and motor power settings can also affect PID. Granted the information is relative and more detailed , its still a lot to grasp. Maybe for the teachers here it helps , but the document I wrote is a simple thing. Of course if people want to know more they now can read this detailed description.
Again , as I only understand JRKS , I cannot comment much further.

Here is some info from when I first did my PID testing on my old motors.

I have changed it a bit since then. I find that would my new motors I can run less D because I don't have to run as much of the "Braking" effect because the motors have less inertia.

With my old motors I ended up going for P=0.5 I=0 D=0.8

I found I got oscillation at P=1.3 and I halved that down to P=0.75 at first but I found that I couldn't add enough D to get it from overshooting.

I ended up dropping the P back to 0.5 and then I could set a D of 0.8 to get it to stop nicely.

With my bigger motors it also seems to help my over all speed having the lower P and D. With my smaller motors I can get a way with running less D so I run them about the same.

I also find that if I run too much P then I get more motor driver errors, this is because I am overloading the JRK, again, if you have smaller motors then you can increase the P because the JRK can handle more gain.

Here are some test I did back on my old motors.

This was at the higher PID P=0.75 I=0 D=1.2. I could stop it on the way up but on the way down it would over shoot. This is stepping the motor from a target of 1000 to 2000 and back again





This is what I settled on with my bigger motors at P=0.5 I=0 D=0.8

Liking the analogy @noorbeast - I have spent a bit of time reading about PID and trying to understand what exactly is going on when changing the values. But as you said, it does get quite involved and some of the terminology went straight over my head when reading some articles about PID controllers.
Even though my setup is running pretty good atm, I'm sure there are still improvements I can make.
Your post definitely helped me to confirm what I thought I understood.

An interesting analogy of a PID controller @noorbeast! But while I follow your logic, I don’t believe the comparison to a moving vehicle is the best. For example, momentum will not cause a car to accelerate past 100 when power is removed at 100 as you stated. Immediately as the power being applied is less than what is needed to overcome the resistance to the car moving forward, it will immediately start slowing down and can never start moving faster (unless maybe you’re on one hell of a downhill grade or driving off of a cliff). Also, with simple P control, it is not true that, “nothing happens until the car drops below 100”. If the car did continue past 100 when its target was 100, P would start applying a harder and harder breaking force the further past the target the car accelerates. And another thing, P control would not be applying full power all the way to 100 and then immediately stop either. As 100 is being approached, P would start reducing power so that when 100 was reached, power would already be at zero. Therefore, when using P control only, it is much more likely that you will fall slightly short of ever reaching the target as opposed to overshooting it unless there is enough inertia left in the system when at the target of 100 to do it.

So while I follow what you are attempting to explain, I’m not sure someone without prior knowledge of PID control would. Therefore, I will attempt to further understanding with a somewhat different explanation (or add to the confusion ).

P, which stands for proportional, is a relationship between the amount of power being applied to the motor and the current error. As noorbeast stated, the error is the difference between the rigs current position and the position that is being called for, the target. As the error increases or decreases, P increases or decreases proportionally based on its setting. P is considered the main control perimeter in a PID system. On a rig that stays in perfect balance at all times, P is all that is really needed to get good control for the most part.

I, which stands for integral, simply consist of a variable that stores the accumulation of all past error values, both positive and negative. It adds the values together and if there is a difference remaining then that value is added or subtracted to P in proportion to the setting for I. In essence, what I does is help P get to the target position faster than it can on its own and all the way to the target position. It should be noted however that I can’t help P get to the target any faster if P is already applying full power on its own. One other thing I can do is help get your rig back to a perfectly centered position also. When using P only, as the target position is approached, P will reduce in power and at some point will not supply enough power to push the rig all the way to the target before stalling. When this happens the error value will accumulate in I until it will push the rig the rest of the way to the target position. However, I should note that you should also set your motors minimum speed to a value just above stall point. Therefore, even if P stalls by falling below this minimum speed prior to getting to the target, the motor will still continue moving to the target until reached.

Many types of systems run perfectly well on P and I control alone and are fairly easy to understand. But D, which stands for derivative, is quite a bit harder to comprehend. But mathematically speaking it is the average of errors recorded over a specific amount of time. These values give an indication of the current direction of control needed in the near future and in effect dampens or brakes the effects of P and I. One example of where D can really benefit a rig though is when it has a heave axis, especially a heave axis that has no counter balance. It this case, the heave motor may never need to actually reverse directions since gravity will already pull it down at 1g. Therefore, without D, P and I would attempt reversing the motor to go downward when what is really needed is to apply less power in the upward direction to go downward.

As far a getting into the process of setting your PID controller though, I’m going to pass on that one.

For a better understanding of D though, here is a link to a more comprehensive explanation.
http://www.controleng.com/index.php?id=483&cHash=081010&tx_ttnews[tt_news]=18861

Thanks. I tend to find that pictures and graphs help a lot. I am not too good with the maths behind it all so I just go by feel and what I see in the graphs.

This has prompted me to have another play around with my PID and I have found a few interesting things.

The main one I really wanted to test was with slow movements like in flight sims. I have found that the scaled feedback doesn't track the target very well with my old setup, it always lags behind on the upwards movement, it never corrects the error.

I played around with the I and it made a massive difference.

This is a screenshot of a steady upwards movement with no Integral. It is like it never corrects for the initial lag as the duty cycles ramp up to the point that the motor starts moving.



Here is the same test with the Integral set to 0.02, it now corrects and follows the target perfectly



Downhill seems to be a bit smoother also. This is with Integral turned off



And this is it with Integral set at 0.02



I recorded a quick video showing the difference. The video really shows the difference well.

How do you set the motors minimum speed on a JRK12v12? I think this is part of my problem with slow speed simulation. The duty cycles need to ramp up to around 20% before it will move the rig if the lever arm are going upwards (if the lever are going downwards then it takes almost no duty cycles), any lower than that then the motors stall. This means it takes a little while for the controller to slowly ramp up the duty cycles so you get a bit of a delay. It would be great if I could set a minimum so it starts at 20% duty cycles instead of having to ramp up.

It isn't a problem with racing games because the error is large enough that the duty cycles jump way up to to a really high duty cycle.

Like this is a example of a fast movement. You can see it jumps straight up to my duty cycle limit, I have it limited to 85% duty cycles. (Ignore the actual curve, I am playing around with different settings)

Awesome thread!!

A good document from Pololu that outlines the PID relationships. Has pictures which always helps

https://www.pololu.com/docs/0J38/5

Section "Tuning the PID Constants"

I'm going to reset my left JRK tonight and try a few things around PID to see if I can work out what is going on with my Motor, as per this thread:

http://www.xsimulator.net/community/threads/need-some-help-sim-not-playing-nice.7656/

Hey guys.

I have done heaps more testing.

Here is a video that I recorded explaining the effect of PID.

This is 30 minutes long but it is worth the watch if you have time.

I start off by showing a few different PID settings while driving and explaining the difference, you can see my PID settings for each test in the top left corner and you will see the pots vs target graph overlayed over the game.

After that I show the effects of each of the PID settings using the JRK tools

Let me know what you guys think or if you have any comments

@Avenga76 - Massive thank you for making this video.

I finally managed to take the time to watch it in full HD and in full screen to watch the details.
I now have a much better understanding of the relationship of P/D, and I also noticed you are using a small of of I.
I've pretty much left that value alone until now, but after setting my P and D to it's tolerances, I set my I to about 0.049999 and it makes a massive difference to the fidelity of the sim.

It seems to really help in violent movements like going up a rumble strip as you corner, you get a real sense of the suspension slapping back down onto the road with the "I" value in play. It certainly feels more focused. It's VERY noticeable in DIRT:Rally where hitting the ground from a large jump gives a nice (and sudden) thud. As you point out, it basically error correction so the Sim more completely follows the onscreen action.

It's not a value that gets much attention on this forum, (I) but is certainly worth looking into. Thank you very much.

My Shoulder mount is almost finished (aesthetics) so I will post a video of it in action soon. (Kids / Wife / Life permitting!)

Hello,
thanks for all the Informations and work you put in this thread and video @noorbeast, @Avenga76.
This helps me alot understanding the whole thing better, and tune my settings for a better movement.
All of you seem to using Jrks, which have a nice software as I can see in the Video.
I'm using The arduino way with Smc 3, where also a nice software from @RufusDufus is avalible.
My question is it enough to turn on the Motion function in Smc3 utilitis and set the motor to following the target position? Or is there a way to do it on the fly while ingame for Smc3?

Ask on the SMC3 thread

That is really awesome news @Archie

I am really glad that my video was helpful to you. I learnt a lot myself while experimenting and recording the video so it is great to pass some of my finding on.