How to control multiple DC motors with encoders 

Sure and thanks for the kind words. Working on another video today as it so happens. :)

I appreciate that and also thanks for stopping by Omer!

@@curiores111 You are welcome, I am a big fan of you!!

Can you clarify what you mean? I've considered that issue - PID with individual motors doesn't seem to account for the initial acceleration phase - meaning that the robot will often turn and then go in a straight line. Is there a way to fix this?

There are several ways to address this. 1) Experiment by trial and error with the initial motor driver inputs, whether duty cycles or PWM, whatever the motor driver takes as inputs to find values for each motor that will cause the robot to run straight. 2) Another way, is to make a series of measurements by running the motors at different input PWM, no speed control, and recording both the time duration, 5 seconds is plenty, and the encoder counts. Using the wheel diameter, calculate the speed of each wheel in cm/s. Make the measurements for a variety of motor control inputs. Plot the results of speed vs inputs. I use Excel and fit a least squares quadratic equation for each motor. Now you have motor equations that relate speed to motor driver inputs. For a selected speed, determine and use the corresponding driver inputs for each motor. Now when the robot starts it will tend to go straight and PID can be used for further control. 3) If you have a good way to measure the robot orientation, yaw angle, use the deviation from a straight line, or the cross-track error, and/or the angle to the goal to make path corrections while the robot is in motion. Do that by keeping one wheel speed constant and varying the speed of the other wheel. John-

Yes, the PID control signal (and consequently the PWM value) is updated whenever the loop function triggers (so the sample time is determined by how often the loop function executes). I didn't use a library, the "SimplePID" is just a class written into the code.

@@curiores111 I did read your code and saw that in loop() you do the PID compute. Regarding SimplePID libary, while I saw in your code that you didn't include any library, there is a SimplePID library so I was confused. However, that I didn't understand that you were using a class highlights my weak coding skills in C++. As a physicist, computations are numerical mostly so DYI Arduino coding is new. Finally, thank you for your prompt reply because my knowledge gained from you is HUGE!

so glad to hear that, thanks Thomas!

Thanks! Yes, it is more difficult to control speed directly. There are a couple of issues to watch out for. I have worked through it before and am planning a video, but just haven't had the time yet.

@@curiores111 Great, I will wait for the new video.

Thanks! I'll see what I can do. 😅

Glad to hear that. 😊 Sure, you can use any standard motor driver. The pololu drivers use H-bridge ICs. That's pretty standard for bi-directional DC motor drivers. Their motor drivers are more efficient/better protected than something like an L298N. :)

Thanks Cilynx. I use Camtasia to generate animations. For the wiring diagram on this one I made an SVG in layers, separated it into several images, and then used crop animations. It's too many wires to animate by hand.

Thanks! I love using templates in Arduino coding, because they allow you to workaround some of the Arduino restrictions at compile time. I haven't used closures to bind at runtime. I'd be interested to take a look. Have a simple reference handy?

@@curiores111 They're really easy in Python. Here's a pretty good explanation: . Python can also use bound methods, which basically captures "self" and returns a callable object that can be used like an ordinary function. Many languages have them. Apple extended C/C++/ObjC in the Clang compiler, calling it "blocks," notably for use with Grand Central Dispatch (threading helper library). I happen to like the blocks syntax, but it is non-standard. Later, the C++ standards committee added closures; I've not used them in C++, so I don't have a handy reference. And I'm not sure if C++ closures can be used where a pointer-to-function is expected (which is probably what attachInterrupt uses).

@@markday3145 Ah, interesting. Feels more like a scripting language construct than something you'd find in C++. Thanks for the reference.

Hi Aoji! Sorry I didn't get to your comment sooner, (long week). So yes no problem if you don't use the leader and follower strategy. You can delete the send and receive functions. You just need to extend all the pin arrays and function calls to four motors (or however many motors you are using). But you also have to have enough interrupt pins available. For the distance of travel to be the same, you need a control loop for each motor (individually). If you use one control loop to control two motors, then you can't actually control the position directly.

Stepper motor position is directly controllable; you don't need a feedback loop since you can simply assign the number of steps to take. That's based on the fundamental differences between how DC motors and stepper motors function.

Hi Osvaldo, thank you for the suggestion! Yes, syncing the motors can be a challenge. Hope that it is working out for you.

So far I haven't been able to identify why some users can't access atomic.h. From what I have read, it's based on your microcontroller. I added an alternative code that does not use ATOMIC_BLOCK -- it's up on the GitHub repository. github.com/curiores/ArduinoTutorials

You can use any encoder motor/microcontroller/motor driver combination. There are plenty of cheap microcontrollers, like knock-off nanos...but I guess now those aren't nearly as cheap as they used to be. As far as encoder motors go, I found it hard to find anything decent for very cheap. Cheap standard motor driver is l298n h-bridge.

Sure, any of those options seem reasonable to me. Two megas would probably be overkill in terms of pins.

Thanks, I've also heard it called "controller"/"peripheral", but I find that a little strange considering that they're both "controllers" and neither one is a "peripheral" (in my mind). For general i2c though, that terminology makes sense. I just got my pico this week! Haven't had a chance to get it running yet though. I also enjoy python (although to be fair I'll code in just about anything).

Thanks Andrew! Certainly, there are plenty of ways you could trigger a start. You could add a remote or RF kit and start when a particular signal is received. You could add a button and start when the button is pressed. You could restart or stop under any of the above conditions. All you have to do is add your peripherals and adjust the control flow of the code to match your use cases.

Of course! I use a lot of different components while making projects like these. My best friend is my 3D printer. It fills in the gaps that invariably arise in fitting the model together (this model's chassis is 3d printed with petg). I like to model components in fusion360, although I recently started doing a few in blender. Usually I'll download the STEP files for the purchased components, put them together in a model, and then add in some 3D printed parts as needed to make things fit together. I usually purchase the mechanical components (gears, belts, lead screws, wheels, and so forth) from pololu and/or gobuilda (metric)/servo city. Generally, if you buy multiple parts from the same brand they will fit together nicely, so it makes it easier to put together a working prototype (and they usually have STEP files so I can build the mechanism before actually purchasing the parts. For example here's the model I used for this build: imgur.com/a/gegIMN9 ). I'll buy cheaper amazon parts if I can find something that looks decent (and need to save money). Also, there are a lot of standard sizes. For example, these are 37D motors which you can find mounts/attachments for from various brands. The motor shafts are 6mm-D shaft which is again, pretty standard. You can then use a hub (or wheel mount) with a 6mm shaft to connect to a moving part (e.g., a wheel). I just added these mecanum wheels ( www.gobilda.com/96mm-mecanum-wheel-set-70a-durometer-bearing-supported-rollers/ ) to this model and I'm excited to see how accurately I can drive it. Sorry about the long reply. I don't usually get asked this question. ;) References... hmm. I guess it depends on what you're trying to build. Anything specific in mind?

@@curiores111 Nothing specific. It's a general, "I wish I knew how to do something like that" moment. I'd like to experiment and learn about affecting the word with a microcontroller (beyond lights and sounds). Something that moves would probably be fun. I'd like to experiment with PID. A balancing robot is probably too ambitious for a first project. I've seen people put a ball in a grooved channel and balance it or keep it at a specific distance (servo with mechanical linkages?). Maybe a fan levitating a ping pong ball, trying to keep the ball at a specific height (though I'd need to contain it so it stays above the fan, not flying off sideways).

@@markday3145 Those are good ones to start with. I am sure you can find projects on arduino.cc for the ball and beam. It seems like people like a two bar linkage for that one ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-FidxDZ7X6OI.html . Very easy to put together. I'd like to see the fan project. My initial thought would be three cheap computer fans arranged in a circle. It would be fun to attach them to small servos to tilt their angle along with adjusting their speed. No idea if it would work to levitate a ping pong ball but it would be fun. (although of course you could just stick the whole thing in a plastic tube and then it would work right off the bat). I don't have any references for these types of projects, some youtube videos and maker projects should be fine. For control theory I like the UMich tutorials, e.g.: ctms.engin.umich.edu/CTMS/index.php?example=BallBeam&section=ControlPID

What motor driver is it? If that's the case, it's probably not bidirectional. If it's not bidirectional, it will be harder to control (you won't be able to reverse directions). You would have to slightly edit the setMotor function to make sense with the control code.

@@curiores111 Its actually known as MDDA 10 Cytron dual channel motor driver

@@Luxigaming Oh well that's good then. Look at the truth table for control logic on the spec sheet. You can adjust the setmotor function so that it matches those conditions. (In other words, you remove one input the setmotor function, and then change the logic inside based on the truth table for a single direction pin, rather than 2).

@@curiores111 Thank you so much for ur help!!

Int b is a local variable for the duration of one call to readEncoder. It doesn't persist past each individual call. The arrays encb and posi persist past this though - they are global variables.

Hi Osvaldo, I an sorry that I haven't gotten back to you yet. I want to look at the code and see if I could identify any possible issues. If the variables are indeed outside their limits, then you would definitely see undefined behavior. However, because the variables are integers, you can perform operations like 0-1, which would give you a position of "-1". This by itself would not cause an overflow. However, if you had pos go over 32,678, that could cause an overflow. In that situation you could consider using a long. If it's not actually a problem with an overflow, it could be a problem with the controller. However, I wouldn't be able to say more without seeing the problem for myself.

@@curiores111 Thank you very much. I understand. Now I'm working on a program to make a robot go straight, because is constantly turning one way, using ESP32 and MPU6050. I hope you take on this challenge with me.

are you driving all four motors with a single driver? For now I'll assume you have separate driver for each motor. There's nothing that guarantees that the motors will run at the same speed. This is one good reason to use feedback control. In general, you'll want to ramp the position slowly enough so that all the motors can get there at (nearly) the same time. Alternatively, you could try controlling the speed directly. In either case, you'll have to make sure that you adjust your speeds slowly enough that the transients are not causing your motors to get out of sync.

@@curiores111 yes, I am using a single motor driver, L298d Motor driver shield has 4 different ports for 4 motors

Hard for me to say without knowing the details of what you're trying to accomplish. I'd suggest starting with a button tutorial (maybe look at OneButton?) and then adjusting the target conditionally based on the state of the button (or intermediate variable).

@@curiores111 What I want is that when I activate the code the motors do not move to any position, until I press the button the motors rotate to the designated position, for example target 2400

@@danielmartinez-mk6iu Sure, so, register a OneButton, "btn". github.com/mathertel/OneButton Then, use a conditional so that if the button is pressed, the target is changed and the PID control is activated. Here's an example pseudocode: OneButton btn(); bool mode = false; setup(){ // register btn // register changeMode function } loop(){ // tick btn if(!mode){ // set motors off } else{ // proceed as usual } } void changeMode(){ mode = !mode; }

Certainly, and thank you for noticing. I love clean code! The way this is set up each motor has an independent PID calculation. To set different rates for each motor, you just set the targets differently. To operate only one motor, you would just create one instance of the PID calculation class. Here, I've put everything in arrays for convenience.

@@curiores111 Thank you so much for your help. You're amazing!

I would but I can't stand that one anymore. If I make a new, improved one, I'll share it.

Hi Andrew! The currT and prevT are recorded in microseconds, so dividing by 1.0e6 produces the time in seconds. So "float deltaT = ((float) (currT - prevT))/( 1.0e6 )" produces the time difference between the two iterations of the loop function in seconds. For me, this is useful as it's easier for me to think in terms of seconds rather than microseconds. The part that says "t = fmod(t,12)" causes the code to repeat every 12 seconds. The fmod function is the floating point modulus (remainder after division). You can find information about it here: www.tutorialspoint.com/c_standard_library/c_function_fmod.htm

Well you need interrupts to read the position... so I would say you can't complete this tutorial without interrupts. But if you have 6 interrupt pins that should be enough. You only need 4 interrupts for this particular strategy. You don't need to use two microcontrollers. Although if you use a lot of interrupts on one controller you might find that you miss some of the interrupts.

@@curiores111 Hi I purchased an Arduino Due which allow interrupts on all pins. Im going to try to operate 4 DC motors off of it using your PID control. I will let you know how it goes.

@@CryingNeutrons Sounds awesome, looking forward to hearing about it.

@@curiores111 Btw I got an arduino due to control 4 dc motors ising your multiple encoders code for velocity control

@@CryingNeutrons well done! I haven't used that myself yet for several motors, so you're ahead of me.😉

Hi Hrishikesh -- that's a limitation of the motor -- it will never run at lower speeds. Generally speaking, if you want to rotate at slower speeds, you would use a higher gear ratio. For example, these 150:1 gear motors: www.pololu.com/product/2828 have a maximum speed of 67 RPMS. They would probably work just fine at 10 RPMS. Here's some information on a wiki page about gear ratios: hades.mech.northwestern.edu/index.php/Choosing_a_Motor_and_Gearing_Combination -edit- Oh I see that I may have misinterpreted your question a bit, and you may already know the answer that I gave. Sorry if that's the case. I find it interesting that you had a different behavior using a pot rather than just a PWM driver. In general lowering the duty cycle of the PWM driver should do the same thing as lowering the voltage across the motor, so you shouldn't need a secondary circuit. In general the most important thing here is the motor and the gear ratio, rather than how you reduce the voltage.

@@curiores111 Wait, I think you got it wrong so when I need to run low steps obviously there is a low PWM signal. So during a low PWM signal there is high frequency which makes motor's coil resonate instead of rotating itself. So I found another way to lower speed. I hooked up a buck converter to output of motor driver which intern is connected to motor and I desolder the pot on buck converter which controls voltage and soldered the emmiter of a transistor to the center pin and connected its collecter to +ve of battery. Then i connected arduino PWM signal to transistor. This way I can run at very low speeds of even at maybe 5 or 6 rpm. But circuit got very big. So I had a question does there exsist such a motor driver which based on arduino's PWM gives a linear DC voltage? Did you get the question?

Resonate in the sense the coil just makes a sound like beeeeeep

@@hrishikeshgawas4579 To answer your question: no, I haven't seen any non-pwm motor drivers. It sounds like you're basically using your buck converter (which most likely has a switching MOSFET) as a motor driver. This is not recommended, as it probably doesn't have the necessary protection against current spikes that can occur if there's a rapid direction reversal of the motor. If the PWM signal is causing an issue, have you considered trying to add a filter (e.g. correctly rated ceramic capacitors) to the output of the motor driver? That should amount to something similar to what you're doing (filtering the output of the switching mosfet), but with safer components. In any case, I'm not an electrical engineer by trade, so that's about as far as my thoughts go on this topic. I think filtering the output and/or finding motors with a higher gear ratio are your best bet.

@@curiores111 Okay I will try filtering the output with a capacitor. Thank you; Yes I agree its not safe but was just trying to prove the point that motor can go on at low speeds but PWM causes problems. I will try it with capacitor and let you know. Thanks 😊.

That's part of Arduino's built in library. No need to acquire it.

@@curiores111 Can only be used with Arduino? Can it be used with ESP32?

@@user-bo3jy8wz5m Sure you can use ESP 32 (might need some logic shifting?). Are you coding in the arduino IDE for the ESP 32? [if so the libraries should still be there]

@@curiores111 I'm perplexed to change the logic from Uno to ESP32.Do you have an example?

@@user-bo3jy8wz5m I don't have an example, but I'm sure there are plenty of tutorials for ESP32 out there. Good luck!

In industry the motion controllers are usually given commands from a PLC( programmable logic controller). Some motion controllers can work by themselves and can be programmed with a series of commands.

it's coming...whenever I find the time... thank you for the encouragement.

да

Indeed, you could.

In industry there is sometimes a need to control over 30 motors. This requires more processing power thought.

I hear comments like this occasionally. It's valid -- not everything is covered, and certainly the methodology is not optimal. But that's the intention -- its not meant to cover everything, and its not meant to be optimal. Generally, my videos are intended to make more complicated engineering concepts accessible to a wider audience. Getting deeper into the details is another step, and not accessible or useful in an introductory video.

error states , " 'recieveLong' was not declared in this scope "

Could you please help me find out where i went wrong? If you could give me your mail id, I can mail the code to you or you may let me know your preferred method also.

You have to make sure "receiveLong" is declared before it is used. In C/C++ you would often declare all the functions in a header file to make sure that this doesn't happen (I didn't turn this into a library, so there is no header with declarations) . The way that I ordered the code "receiveLong" is declared before it is used. My email is curiores@gmail.com if you'd like to email something.

no it should be long, because it returns a long.

Hmm i get a error stating the recieveLong() isnt declared in scope for a statement under void loop(). Anyways i have mailed my code. 😊

Thanks for the feedback Daniel (albeit a little rude ;)

Yes, you can... but I think there may be some issues with slow interrupts because there is a lot going on in a raspberry pi. I am not familiar enough to know how well it would perform, but I'm guessing not that well. I would myself prefer to do the control loop on a microcontroller and then have the pi doing other tasks, like computer vision and path planning.