PID Line Follower for EV3 - The Ultimate Line Follower!

[@williambruneau8667]

Thanks a lot, I am programming in C++, but only clearly understood the programming when searching for your video. Great job

[@TheToadcode]

Thank you for all your hard work. I teach the concepts to my FLL team then I send them, to your site as follow-on. You are indeed passing on the knowledge to a younger generation!!! KUDOS

[@Builderdude35]

Thank you! I am honored to have had such an impact!

[@jmeng06]

Thank you for the help! I have used this as a baseline to understand conceptually how to code a PID control using C++! Before, I didn’t understand how this worked, but now I know how the math works out. You have been a great help and allowed our VEX robotics team to go to State Championships due to us heavily using PID control for our autonomous😅

[@technicalwaffle9859]

Really nice video! Helped me a lot on FLL! I also used this PID feature for our/your Gyro following program that we are also using in FLL. The Gyro follower was only proportional, but now it can also use PID. You can probably make a tutorial out of this too. You really don't know how much you've helped me and my team in programming.

[@Builderdude35]

I'm glad to hear that I've helped! And thanks for the suggestion!

[@omsrijan]

thank you, this line following video was really helpfull for my FLL team

[@CH-fh2lh]

Great video. I am a robotics teacher and am rolling out an EV3 based robotics class. I could not find good documentation on the EV3 block language. This video answered a lot of my questions and saved me a ton of time! I am going to try applying the PID to a standing robot. Rather than the gyro sensor I am using an ultrasonic to measure the amount of tip. A proportional loop is insufficient. Hoping the I and D will do the trick. I am worried that the problem is more the rate at which the program runs. If so, I guess I will have to learn Python

[@acutosan7335]

Find the Kp value first. If u want the robot to correct the error fast then a higher Kp is needed at the cost of some jittery during the line follow. If u want a smooth line follow then a lower Kp shall be applied at the cost that the robot cant follow the line when coming from a steeper angle ( doesnt correct error fast enough). Ki should always be 0 unless youre using a gyro sensor and wants to go for a curve. Kd kinda supports Kp so trial and error is the way for me. the first value i go for Kd is according to my Kp. So the lower the Kp the lower my first guessed value for Kd. its the most effective for me

[@harygren4033]

you are very clever and can explain very clearly. Thanks ! I wish my teacher at the uni explained it so well 20 years ago:)

[@Builderdude35]

Thank you! I am honored!

[@ryanberardelli2363]

Great video and very helpful, haven't tested it out yet, but this was a big help on my college project. Thanks!

[@Craftymango5]

Hey

[@Craftymango5]

Can you help me I can’t do it.it is to hard

[@meltemerzurum742]

hi! i do the same things but my robot doing like zig zag or doing nothing except go straight what can i do

[@mostafashabanit]

I need the programming file please . And the structure of the building

[@neilorion9953]

Shouldn't the K values vary with the speed that you are going higher K as faster you go, or are the K values independent from your speed?

[@honeybanga2865]

the same issues mentioned before. Tried different values for Kp, Ki and Kd. The robot follows a straight line perfectly when Kp is set to 1 and Ki and Kd are set to 0. Then, when I try to play around and fine-tune the program, the robot goes off. It follows a straight portion of the line well, but when there is a turn, whether sharp or round, it just goes off. as I am in the wro team I need to use 1 sensor for line following

[@Builderdude35]

If you are having too much trouble adjusting the PID line follower, I recommend changing it to a proportional line follower for now to make it a little easier to work with (set Ki and Kd to zero).

[@KIXEMITNOG]

Please make Arduino robotics videos as it would be great and very helpful. I suggest this because you are the most helpful youtuber.

[@Builderdude35]

That's an awesome idea, I will consider it.

[@KIXEMITNOG]

@@Builderdude35 thx

[@pdmarsh5466]

Excellent tutorial. Thank you very much. I'm not sure that I have PID tuning down, though. It seems that proportional control is pretty much all you need when going in a straight line. I found that Ki and Kd can remain at 0 while still getting smooth and straight travel. However, you do start to get waggle in a curve. I decided to tune while the robot is on a curve. Even so, I don't think I really understand how to tune a PID loop.

[@jennipickett]

Thanks! You are a great teacher!

[@austinharris5700]

For the people who have their robot spinning in circles when any Ki value is used the fix is to put another write integral variable set to 0 in the loop, It only works if the variable is at the start of the loop or the end. The reason that this fixes it is the integral is getting higher and higher as the program runs and makes the robot go in circles I hope this allows the program to use Ki. :)

[@morgangadd4988]

This does fix the problem Thanks

[@Builderdude35]

The problem is, this effectively removes the integral term; it's just the same as if you had just set the Ki to zero. Then, what you really have is just a PD line follower.

[@cronicas_imemoriais]

And It doesnt Go higher and higher, It gets higher when the error is positive, but the correction and the negative setor Will balance It almost at 0, you are probably using a Very high ki value.

[@wackings7776]

Can you give me the link to this program?

[@404nohandlefound]

You can also make the threshold percentage to be adaptive as well, making this code impossible to go haywire, this allows it to work in dynamic lighting conditions.

[@Builderdude35]

Great suggestion! Thanks for sharing!

[@andychu2532]

how do you do that?

[@ob4394]

Hello do you know how to calculate the conversion factor for ev3 robot to spin in place ??and how would I turn that into robot degrees?? I’m struggling in robots and teacher not helpful.

[@jerrymk6846]

I have a problem, I think the sensor value is not linear to the offset from the target. What I mean is in the proportional part, you do value - target as the error, but let's say your robot is exactly on the edge of the line. If the robot shift for example 1 mm to the white side of the line, the error will not be same as shifting towards the black part 1 mm. Will PID fix it? I don't think so, so should I calibrate both side and try to calculate the none linear formula to convert error to robot's offset or it is too minor and is not an issue?

[@Builderdude35]

A PID will not help with this, because PID also assumes that your system's error is linear. Calibrating both sides (by measuring values on both sides at various points and creating a regression) is a good way to go.

[@jessnam]

holy shit its 2 am and u just blew my mind thank you!

[@laurenreynolds4132]

can you make a video of this code commented?

[@ScottMorgan88]

Awesome job. Well done!

[@advaitpatel9940]

Need help! I've tried to program this line follower, but it doesn't work. Instead of following the black line, it just turns. I took out the Ki and Kd values but still it didn't work.

[@omeghana]

I tries it but mine was in spin loop.Can u help me.Thx.

[@fabiolally]

Mine too, it just spins nonstop

[@reid-dye]

Mine did too. Just turn your Ki way down

[@benciruiz6493]

Hello, can somebody explain to me how the robot go on the right side? I didn´t quite understand how to do so in the video.

[@ncf3535]

How do you go along tuning PID? tune KP Then Ki?

[@Builderdude35]

Yes, Ki should be the last to be adjusted.

[@yanganzhi]

can provide 3 sample kp, ki and kd values for us to start testing from?

[@吳峻霆-f1v]

HOW can I GET TO THE INTEGRAL?

[@2nightwehunt]

Builderdude? More like Buildergod

[@jeftemelo4452]

Boa noite Sou do Brasil e começei a estudar programação vc poderia me dá alguma dica ou disponibilizar essa programação para eu aprender mas sobre PID

[@jeftemelo4452]

Tenho um pouco de noção sobre o EV3 qual vídeo , seu você me recomenda assitir ?

[@pieterpretorius1014]

how can this be adapted to control the motor like a hobby servo motor that only move within a certain range? i need that kind of behavior for the motors because of an idea i have so i can build a helicopter swashplate demonstrator

[@Builderdude35]

You can program the EV3 motors to rotate a specific number of degrees (or rotations). That will enable to control the rotation to do whatever you want.

[@imarrb809]

kyle, you missed the part where you put to go until color sensor senses 35

[@attupsgeorge224]

Thanks for sharing your videos. They are very useful. How do we change the target value of the program at FLL tournament since we cannot edit the program with a laptop. Appreciate your suggestion

[@timo5727]

You can edit during practice runs on practice tables.

[@joseclaudio568]

Excellent! Very good indeed! Thanks for everything you taught me, I can not wait to test and calculate my variables!

[@Builderdude35]

Fantastic!

[@rochesterrobotics5259]

Why do you use a target value of 35 instead of 50 (half dark and half light)? Would the target value need to change with each re-calibration at an FLL tournament?

[@Builderdude35]

Ideally, the black line returns a reflected light intensity of 0%, and the white line returns 100%. However, reality almost never fits this description, because the light intensity measured by the sensor varies based on 1) the ambient light conditions, and 2) the surface the sensor is observing. In my case, the black line returned 5%, and the white line returned 75%, so I use 35%, which is about halfway in between (and directly measured from my target).

[@laksshaysehrawat5559]

can we make a line follower which does not need any calibration. calibrate once and it will calibrate the values further by itself. Please tell and if possible make a tutorial.

[@Builderdude35]

It's absolutely possible. In fact, none of the line followers I have made tutorials on require calibration, nor have I ever had the need to run a calibration.

[@joseespejoaliaga4053]

you can give a demonstration of the program on the robot?

[@Builderdude35]

Coming soon... Watch out for next week's video!

[@hero_16-f1w]

For derivative isn’t your last_error always going to be greater than your current error so derivative is always negative. That doesn’t seem right. Am I missing something. Can someone please explain? Thank you

[@Builderdude35]

There isn't anything that says the last_derivative is always larger than the current error. Ideally, we would want error - last error to be less than zero because that means the robot is trending towards the correct path and will eventually reach zero when both error and last_error = 0. However, the robot is in a feedback loop and will stray off the line as the line curves or because of other forces that cause the robot to deviate (ex: imbalance of friction), in which case, last_error > error and derivative > 0.

[@hero_16-f1w]

Thank you it makes sense now. Great video btw

[@mariosnacouzi8180]

Does a pid line follower get you much better results than a proportional one. And would it make a difference in a wro competition?

[@Builderdude35]

A PID line follower will get you better results only if adjusted correctly. It takes a while to get a PID adjusted correctly, while you can get a proportional line follower up and running in a minute. You'll have to weigh performance with time requirement. In most cases, a proportional line follower serves your needs just fine.

[@NikolasSamaras]

I would like to ask what is the range of values of Ki and Kd. Especially Kd can be over 1 or always between 0 to 1?

[@Builderdude35]

The values will be specific to your robot, so I can't necessarily tell you exactly what they will be. I will sat that either Ki or Kd -- I forget which one -- will need to be VERY small.

[@countdownquest]

In my tests usually the Ki value is VERY small in my gyro following program.

[@GiorgosPamboris]

Start by Kp=1 and experiment until your robot follows the line. Have Ki=0 and Kd=. Then slightly increase Ki=0.0001 and see if it gets better. Then Do the same with kd

[@Aeroheath]

Is there a line follower that stays on the line that the color sensor doesn't go on the line half, but actually stays on the line?

[@Builderdude35]

You will need to use two sensors for that. See my tutorials on how to program those.

[@namay7449]

You have an extra block on the green tab(Action tab), can you tell me what it is??

[@Builderdude35]

Thant is for the Mindsensors GlideWheel, which I used for my Thundersmart IV project. They are third party hardware that essentially turns a Power Functions motor into an EV3 motor. You can find it on the Mindsensors website: mindsensors.com

[@percyho7013]

what is the use of the last error?

[@apriljoyalbior5809]

Great help!

[@jasonfath6948]

Could you use 2 sensors to make it follow straighter?

[@Builderdude35]

Indeed. I have a tutorial on this.

[@roboticsc7053]

I tired it, and it is working. However, it only turns right and does not turn left when suppose to do. I don't know why can you help with that please? and can you please explain how to do it for more than one sensor?

[@Builderdude35]

The robot should be able to turn in both directions. I recommend adjusting the first value you subtract from your light sensor reading.

[@jaderyandelrosario4846]

Builderdude35 We also have the same problem.. What do you mean? Can you show us a video that shows the trials?

[@roboticsc7053]

Actually I fixed it by changing the target value. You have to put between the black line and the other color you do have. Then, measure the go to pot view, check the value there, and get it to be written as a target. After that, try to fix the Kp, Kd, and Ki again. In my situation, I have the target = 55, Kp = 2, Ki = 0.00001, and Kd = a big value but not fixed yet.

[@quipoxd72]

i try your coding but my robot is turning around in place. Virtual Robotic Toolkit

[@French28Pilot]

Hi i know its 2023 but my robot is always spinning and when i put it on the line it spins aroun itself. How to fix please?

[@bangerMURILO]

ajusta a luminosidade

[@Crunch_Games]

I tried to add a Ki and it did not work at all but when I took it out it worked. What is the problem and what should I do?

[@Builderdude35]

The Ki value should be very (read: VERY) small. Experiment with it for a little while. If you still don't like it, set your Ki to zero, which will effectively remove it from the program.

[@olfmombach260]

Keep in mind that with this edit you delete the whole Integral thing out of the program, since you multiply the Integral value with zero, so it is not just the K(i)-value that is being crossed out.

[@Mark-jj1fc]

Hi! We tried to use the PID line follower... We calibrated our sensor 1(minimum, black line colour intensity) to 100 (maximum, white line colour intensity). We looked at our Black-White midpoint and it showed us an ,,n,, value. We put the value into the Midpoint value and started the program. This did not work.. the robot started to go circle around ... We tought that the Kp value is too big, but after we modified the value the robot did the same thing. What is the problem.. Please answer earlier as you can.

[@Mark-jj1fc]

FIrstly, I am using the normal PID line follower with just a Kp valu.. Secondly I did not try to set my Kp value to 0. I think taht this program is not working because I can not set the middle value (50% white 50% black).. I calibrated but this is not working.. How can i set a constant value for the middlepoint?

[@Mark-jj1fc]

I put at the middle ot the white and the black and i saw a number ... After this i wrote that number into the program and it did not even work.

[@jaderyandelrosario4846]

I got the same problem

[@timshort2601]

t job explaining keep up the good work

[@Rxnkshitij]

are you sure this will work ? Cause i have WRO in one month? How to measure ki,kd,kp values??? on what factors?

[@Builderdude35]

The PID line follower takes a lot of time to properly calibrate. If this is too much time for you, don't use a PID and use a proportional line follower instead.

[@Rxnkshitij]

on what factors and i ahve got a lot of time

[@Rxnkshitij]

I have made the program but is there any way to make just PD llne follower

[@Builderdude35]

That video is coming very soon!

[@michaelyang1459]

I dunno how to know if my values are getting better or worse

[@milesrothman4345]

I don't understand what any of the variables have to do with anything. what are they for?

[@Builderdude35]

Storing, transferring or reading values. Each one is explained in the video.

[@pavelyarovoy]

Your set of videos on PID is fantastic, thanks!is it possible to apply PID to two sensors to achieve the best result in a competition? If not, then what is better: 1-sensor PID vs 2-sensor proportional?

[@Builderdude35]

Thank you! It is possible to apply PID to two sensors, but there's no need for it, because it's excessive. 1-sensor PID is very difficult to refine, but if you have the time and patience, very little can match its performance.

[@ncf3535]

Is it possible to have just PD instead of PID?

[@Builderdude35]

Yes, I recently found that PD is better than PID... I'll have a video up soon that explains why.

[@pbmann69420]

Is this overkill? Or would proportional line following work fine for WRO line following purposes?

[@Builderdude35]

Yes, this is overkill. A proportional line follower will adequately serve your needs and allow you to bypass a lot of the time associated with adjusting the PID line follower's K values.

[@Brian-jn2pi]

Peter Bergmann My team in FLL used a PID controller because we could go so much faster than just a P controller

[@gregcollins3404]

Do you find that doing a complex algorithm slows down the control loop so much that a simpler, faster routine actually does better? Have you ever seen or found specs on what the execution speed of the EV3 is? People that are having trouble getting a line follower to work its often because their speed is higher than the line follower can respond to.

[@Builderdude35]

I'm not sure what the actual EV3 computing speed is, but I would say that this line follower, as complex as it is, will not significantly slow the EV3; I have run much more complex programs on the EV3, and it seems to take a lot more complex programming before the EV3 slows down.

[@gregcollins3404]

It seems like on my EV3, it is very easy to go through a line and off the other side before the EV3 responds to it. We have to keep the speed pretty slow to stay at all smooth...

[@meganton9417]

Builderdude35 i tried to solve this by putting the single parts in differented loops and i have the feeling that it works better now

[@hutai66]

This mans is the biggest brain alive!!!

[@yanganzhi]

Loved your video... Could you provide code for download though? Thanks!

[@aznbreaker37]

demonstration of the line follower in action?

[@Builderdude35]

I'm going to have a line follower comparison video for next week, which will show all of them working back-to-back.

[@leyawonder2306]

your channel is the best

[@blu3mo]

Is it possible to download this data?

[@Builderdude35]

Download what data?

[@blu3mo]

Data of the program (ev3 file)

[@Builderdude35]

No, the program is not available for download. You will learn everything you need to know from the tutorial.

[@geneworxoughtamation372]

Coach here. BuilderDude -- You may not recall, but a couple of years ago but we suggested that you do this tutorial after a team member struggled to implement PID on our robot. You've done a great job on this. While the team implemented this a couple of times over the last few years, your explanation is through and clear. I used this in an advanced FLL/Robotics class this summer and the kids (12-14) were able to successfully implement it. I also then had them convert the program to work with the gyro instead of the light sensor. That worked well as well. I now have a couple of suggestions for followups to the PID video/suggestions for more advanced videos: - Look at this video (goo.gl/5hoZFq ) which explains the PID in terms of how it adapts to past and expected performance. - You might make a follow-on video showing how to implement the gyro PID (very useful) using the explanation of past/present/future to inform your discussion. - I see you just posted a replacement line follower tutorial. You might do the PID update along the same lines, incorporating the Gyro into the video as well/instead. You might show how you can integrate "warp speed" with the PID controller to have the robot accelerate and decelerate to the destination as an add-on to the PID by making speed a variable that's controlled by a My-Block. - You might show how to use both the light sensor and the gyro sensor at the same time to follow the straight line more accurately by using both the angle and light intensity. (Hint: use a Kalman filter for this. Unfortunately, Kalman Filters are best done with matrices, and LabView doesn't handle Matrices well). I mention this one in part because I expect you are getting more advanced in math, and might enjoy learning about Kalman Filters if you liked the PID. - You might do a survey of characteristics of the best-scoring robots in at world championships (sensors, motors, size, attachment mechanisms, attachment types, etc). However, I understand you might just be ready to move on to other challenges. Either way, you've been a great asset to the FLL community over the last couple of years, and you've been an inspiration and positive impact of at least tens of thousands of youth around the world. Keep up the outstanding work. If there's anything GO can do for you, please let us know.

[@Builderdude35]

Thanks for taking the time to share this! It's almost as if you've read my mind! I already have a PID follow-up video in the works, which goes into more depth about what each term means, just like in the video you included. It will be out in September. I will also look into getting a PID gyro video out for October. As for the Kalman Filter, I love the idea, but I think it's a bit overkill for something like this... especially considering the EV3-G programming environment does not do matrix operations (at least not to my knowledge). I am currently using a Kalman Filter in a personal project of mine, which involve GPS and is programmed in C. Anyway, thanks for sharing your thoughts!

[@lukeqiao5477]

can you demonstrate the PID line follower please?

[@georgegeorge690]

is this better than a two line proportional? if yes is it possible to make a two light sensros pid?

[@meganton9417]

George George of course its better just because of the less sensor port because you can coose better where on the line you want to drive along. Youre second question is a bit difficult because when you have two points on the line and want to move them you have one point Besides the line where you cant control it

[@georgegeorge690]

Meganton thx

[@404nohandlefound]

This video is awesome, I have one question, you keep adding error to integral, wouldn't that increase the integral value to a lot eventually ?

[@Builderdude35]

Yes... This is something I will discuss in a video that I have coming soon.

[@izakkanter452]

Yes, and if the line has a very sharp curve, your robot will make a very wide turn. This turn becomes less wide thanks to the integral. However, when the robot finally reaches the line, the integral value is so big that the robot overshoots the line. When the robot crosses the line though, the error value becomes negative, and you thus begin to subtract from the integral. The robot oscillates around the line until the integral becomes zero. Questions?

[@elipregerson9065]

Error can be both positive or negative, depending on which direction your robot is off of the line. These two possibilities work to neutralize each other over time so that the integral does not get too big.

[@gismosgadgets8343]

well done

[@arhamjain7306]

It totally confused me but I will make the program and I think I will understand soon.😅😅😅😅

[@cksmaya]

Think you very mach.

[@GuyEvron]

Great video. very helpful !

[@Builderdude35]

Thank you!

[@lancerafael8975]

Anyone here knows how to do this with raw sensor?

[@lukeqiao5477]

nice video! it helped me a lot!

[@ranveerj5021]

If integral is something, but error finally becomes zero, it just adds zero to integral but integral does not change it remains the same value it was before. So it multiplies to ki and it will correct, even with zero error!

[@jeudygamboa7390]

I don't speak so much English I'd like you show what the robot does after programming. Greetings

[@Builderdude35]

Maybe ou can send me an email through my website, builderdude35.com?

[@Cloonsey_]

Epicly

[@pbmann69420]

It would be great if u could show us this in action

[@busteman]

Use this link: kzbin.info/www/bejne/hmaTdHhmrdybrNE. Near the end is the pid

[@chandrimatitk6250]

I love robotics and how are you

[@danielkatanski6503]

Good programming example. You are influencing programmers young and old. Will you please improve your program examples by adding comments! In my FLL team students put in comments before the code blocks; if you cannot briefly write the intent in words, then the odds of getting the code correct is greatly reduced. Lead by example by documenting the what and why of what your code is doing.

[@arefhafza4070]

Can show how we can save time during fll rounds

[@Builderdude35]

See my video on Dog Gears. That's a good first step.

[@rishiveerepalli4957]

How do you get the target light value? . . I didn't understand that part

[@rishiveerepalli4957]

Never mind. . . The tutorial on the proportional line follower explained it.

[@armanbhat891]

Me seeing the 500th variable block being added : "NOOOOOOOOOOOOOOOOOOO!!!!!!!!!!!!!!!" Jokes aside this is a very useful video

[@spazywazi8673]

I have tryed most of your coded and they all helped me a lot during our FLL season... I am now working on the PID program but not using light sensor using Gyro Sensor to help keep robot move in straight lines. Do you have an idea how to do this PID ?

[@Builderdude35]

You are more or less going to use the same program, just replacing the color sensor with a gyro. The rest of it stays the same.

[@spazywazi8673]

I have tried to do so ... but i got confused with callulating the error of the gyro reading where the goal i want the robot to move on is 0

[@gevink6174]

Great video, but since I am in the Robocup team representing Melbourne,Victoria, on one of my practice runs, instead of following the line the robot just went crazy bashing all the other robots making me one of the worst!

[@Builderdude35]

Oops! I'm sorry to hear that. I guess my best advice would be to always make sure your program and sensors are optimized (calibrated) for your environment.

[@lukeqiao5477]

is it possible to build a two sensor PID line follower?

[@MichaelTheGuy]

yeah

[@roboticsc7053]

can you help with the two sensors?

[@danieliorio818]

How about gaps?

[@Builderdude35]

I have a separate tutorial for that... check it out!

[@danieliorio818]

@@Builderdude35 but is it like a complement for this video? Can you send me the link?

[@Builderdude35]

kzbin.info/www/bejne/gonHnmljfNFjjac

[@rhianngacusan1227]

nice!

[@LucEv3]

Awesome!!!

[@gurmeetkaur-vw2ry]

instead of kp can i use the advanced math block and set it to a-b*c and BTW great channel

[@Builderdude35]

Try it and see what happens.

[@veebhushah9667]

How do you condense to my block

[@Builderdude35]

The same way you would with any other program, but you will have A LOT of parameters to set up. Check out my MyBlock videos: kzbin.info/www/bejne/mH6un3qja66AbM0 and kzbin.info/www/bejne/b6Gym4yNmbKig9k

[@VINICIUSSANTOSDELIMA-g2i]

14:55

[@perdredupoids-vivresaineme9564]

Mine is just spinning

[@user-pk4ut1ge3t]

it doesn't work with me i'm helpless

[@رفاهالسقاف-ث8ز]

I wanna program this in NXT 🙂💔

[@kaandere5833]

please make a video a test it

[@Builderdude35]

I'm going to have a line follower comparison video for next week, which will show all of them working back-to-back.

[@kaandere5833]

thanks

[@charmingotter976rr]

That’s not builderdude35 that engineerboy53

[@Stable_Robotics]

Hello builderdude! I found a video : kzbin.info/www/bejne/g3nIpIGNgrCHhrc&ab_channel=Valter Is this based only on PID control? Does it need any arrays? Thanks!