What Is Sliding Mode Control?

[@MATLAB]

If you have any questions or comment, please let us know!

[@HansScharler]

In the video, you mentioned that sliding mode control can be used to control a robotic arm. What are the specific challenges of using sliding mode control for robotic arms and how it can be addressed?

[@WailTheDesigner]

I would like to understand the difference between regulation, tracking, and servomechanisms. Does regulation mean that the reference is constant, and tracking that the reference is variable? How does the servomechanism fit into this?

[@Jair_inacio_Neto_Teixeira]

This simulink block is only available in matlab 24?

[@BrianBDouglas]

@@Jair_inacio_Neto_Teixeira Yes, the Simulink block just came out in 24b. There are examples implementing SMC in MATLAB from earlier releases but they don't make use of the SMC Simulink block.

[@BrianBDouglas]

@@WailTheDesigner in the case of the SMC block in Simulink, regulation mode drives the states to zero. And tracking mode drives the states to a reference. In tracking mode, that reference could be constant, variable, or even set to zero (in which case it behaves the same way as the regulation mode). Around 17:00 in the video, you can see how the switching function changes as I toggle between regulation and tracking. A servomechanism is something different. It is a mechanical device whose position or velocity is controlled through feedback. Probably most servos are controlled with some form of tracking mode since you are often asking the device to move to a given position or velocity. Does that answer your question?

[@khantalks0]

Brian Douglas Godfather of simple explanation of control theory

[@MATLAB]

He is amazing!

[@BrianBDouglas]

🥰

[@farzanehbagheri6379]

One of the best explanations I have seen from SMC.

[@MATLAB]

Awesome, glad it was helpful!

[@Erik-ii5rc]

Agreed!

[@nurahmedomar]

I have been working on SMC for the past year, and after watching this video, I understand it even more clearly.

[@MATLAB]

Thrilled to hear that it was so helpful.

[@walterp773]

This is simplicity. I feel the concept was extremely easy to understand when the vectors appeared.

[@MATLAB]

Yes, vectors makes things very simple.

[@annad9794]

This class nearly drove me crazy this semester, but your clear drawings really saved me!

[@MATLAB]

Glad to hear it was helpful!

[@BillTubbs]

I think the boundary layer function at 16:08 might have a minor error. Should the third condition be s < -phi (not s < phi)?

[@BrianBDouglas]

Yes! You are correct. There should be a minus sign in front of phi. Thanks for being so attentive. I can't fix the video anymore but, hopefully, I can find a good way to call it out so that it doesn't confuse anyone else. I appreciate it!

[@BillTubbs]

@@BrianBDouglas No worries, it's quite obvious once you think about it.

[@Jair_inacio_Neto_Teixeira]

Simply amazing! Please do more videos on non linear control

[@MATLAB]

You got it!

[@sasankpotluri4422]

Hands down, one of the best explanations on SMC. Felt like a 3Blue1Brown video.

[@MATLAB]

Wow, thanks!

[@MrHaggyy]

Awesome explanation from Brian. One might say as usuall at this point. I used SMC on a speed control of a system where mass and resistance have high uncertainty and will change quite a lot during operation. Might be interesting to show how uncertainty in the model can break traditional controls like PID, but not SMC.

[@MATLAB]

Great feedback, thanks!

[@BrianBDouglas]

Thanks for this comment. I thought about trying something like that in this video but it was already 20 minutes long! PID can be pretty robust though, but often just not as robust as SMC. Are you still building systems with SMC? What do you find are some of the challenges with it? Thanks!

[@MrHaggyy]

@@BrianBDouglas Thanks for the reply. Yes, keeping explanations compact really is an art. I haven't touched it in a while. We prefer using PIDs when sufficient as more people know how to work with them. But it's definitely a tool if model parameters change a lot. Shattering was challenging. It can work the actuators pretty hard and introduces vibrations into the system. You need to be aware of them and how they might affect your sensor readings. Likewise precision is a tradeoff with the reaction to disturbance. If you make the boundary layer bigger and reduce the gain you get more error from disturbances. If you make it tighter you recover faster, but get more shattering. Having some friction in the system did help with shattering as the actuator couldn't follow as much. An integral part inside the boundary layer can reduce shattering. A term that scales the gain depending on how far away from the sliding mode you are helps with the urge of the system to shatter when it reaches the sliding mode. I'm looking forward to the next video.

[@achmadsyahrulirwansyah5304]

Finally, you explain this. I've been waiting Brian Douglas to explain SMC for me. Thank you.

[@MATLAB]

Yay, glad it finally happened!

[@harrytsai0420]

Great Video from Brian Douglas!!!

[@BrianBDouglas]

Thanks!

[@fea6849]

Mr.Douglas, you are awesome!

[@MATLAB]

Yes, he is amazing!

[@yahyahaque1850]

Best ever explanation I watched. You are a true gem

[@MATLAB]

Thank you!

[@bachaaymene]

I would love to see more video about the terminal sliding mode, super twisting ...

[@MATLAB]

Thank you for your suggestion!

[@351Braeden]

It would be great in the next few videos to show an implementation of SMC for temperature control. A lot of temperature control systems use PWM to control some heaters output so a continuous control value (duty cycle) has to be converted into a discontinuous control signal anyway. It would be awesome to see to computation and implementation of the discontinuous control directly!

[@MATLAB]

Great suggestion! Thank you.

[@Pedritox0953]

Would be great a step by step what the control does in time. Great video!

[@MATLAB]

Noted, thank you for your feedback!

[@ChibunduUmeh]

GOAT of matlab control theory. Can you do a basic video on the use of functional analysis in control theory? I always understand more complex concepts better after watching your videos.

[@MATLAB]

Thank you for your suggestion!

[@shyennepinheiro4547]

simply fantastic and very educational

[@MATLAB]

Thanks, glad you find it helpful!

[@xlittletomiq6777]

Great video, thank you! Could I have two questions for further clarification? 1. Regarding the “boundary layer” conditions.. I assume that theta=-1 iff s < -phi? (Emphasis on the minus) 2. Solving the equation for u(t), the output is a 2x1 vector. What is the interpretation of this and which element of this vector is then used for the control?

[@BrianBDouglas]

1) You are right! Great catch. I forgot the negative sign on the phi. Thanks for the sharp eyes! 2) u(t) should be the size of the number of inputs into the system. In the case of my example in the video, u(t) should be 1x1. Can you explain how you're getting 2x1? Maybe one place for error is with f(x). f(x) is A*x and not just the A matrix. So, A*x is 2x2 times 2x1 which makes f(x) a 2x1. C transpose is 1x2 and so 1x2 times 2x1 is 1x1.

[@xlittletomiq6777]

2) you are absolutely right, I did not realize that f(x)=A*x and not only A.. Thanks for the quick reply :)

[@jeromelachaize6959]

Realy nice explanation of SMC

[@MATLAB]

Glad you liked it

[@alidoraghi3852]

May we use gain scheduling technique to chattering reduction of smc control?

[@SimonNjuguna-f4f]

How is the disturbance factored in in the implementation?

[@CH-qn9mc]

Is it possible to use neural state space model as the system? Great video!

[@oldcowbb]

finally, i have been hearing sliding mode control for so long but didn't bother looking it up

[@MATLAB]

I hope it was worth the wait!

[@oldcowbb]

@@MATLAB well worth

[@simaoalpha1198]

Great video as always, although It would be preferable in 12:20 to explain stability in the sense of Lyapunov rather than the graphical explanation.

[@MATLAB]

Thank you for your feedback!

[@MrHaggyy]

True, but i get why they choose a graphical explanation in a video.

[@BrianBDouglas]

Thanks for the feedback. I wanted to try to explain it in a way that maybe isn't often done. I figure that there are already Lyapunov-based explanations out there and instead of repeating them, I was trying to come up with a complementary explanation.

[@AvelinoTiago]

what you mean by preferable? literally the entire internet/literature is about lyapunov.

[@official-michael1993]

It was helpful video. I'm working on advanced SMC for application of nuclear reactor power tracking and disturbance rejection princples. SMC is usefull for controlling Xenon oscillation during load-following operation of nuclear reactor, however to control chattering effect, we can add fuzzy algorthm methods.

[@MATLAB]

Thank you for your great tip!

[@hasinabrar3263]

Brian Douglas the GOAT of control;

[@BrianBDouglas]

☺️

[@emmanueleyi913]

Can SMC deal well with time delay system?

[@BrianBDouglas]

I think in general it can handle some amount of time delay due to the fact that there is some robustness built into the controller, however, I don't think the standard SMC controller is designed specifically for time delays. For example, a delay in the actuator would immediately increase the chattering around the sliding surface since it would take an additional amount of time before the opposite sign is applied to correct it. I haven't looked into SMC explicitly with time delays so hopefully, someone else can chime in if they know.

[@tigerwuli2760]

shortcut at 16:52 please help

[@MATLAB]

If you need help with the math, please post your questions in MATLAB Answers www.mathworks.com/matlabcentral/answers/index

[@denissopichev5986]

What the difference between Sliding Mode control and lets say Adaptive gain which depend on error size for example? Its basically the same from my perspective

[@BrianBDouglas]

They are both robust control strategies, but they approach the problem slightly differently. SMC has this discontinuous control where it switches back and forth to slide along the switching surface. As I explain in the video, this high gain switching can allow for the system to be really robust to parameter variations and external disturbances. With adaptive control, the control gains are adjusted smoothly and continuously to account for changes in the system dynamics. I don't believe adaptive gain is as robust to uncertainties and disturbances as SMC because of the time constant involved in adjusting the gain. Adaptive gain doesn't have the chattering issue like SMC has so there is a trade off as far as I'm aware.

[@mayazhussain]

100 times better and easier than reading those pesky papers on SMC....

[@MATLAB]

Glad you like it! Check out other videos by Brian Douglas.

[@FatherGapon-gw6yo]

This is super-doing some NDI controllers for an aircraft and chatter is bad-should be able to do the same gain trick.

[@MATLAB]

Glad the video was useful for your current work!

[@p.s.l.7763]

14:45

[@arrijalrifai2197]

I hope you create this video 4 years ago 🤣 This topic was burn my head 🥲

[@MATLAB]

Hahaha, better late than never!