LQR Method (Dr. Jake Abbott, University of Utah)

[@tylerhamer]

Wow. This was great! None of my professors at MIT have been able to teach this physical intuition. Thanks so much!

[@smilesmile787]

+Tyler Hamer yes, his example was very straight forward

[@mkhalil007]

OoOo

[@venkr1728]

Brilliant! A good teacher explains the complexity of a problem, but a great teacher simplifies a complex problem. This is an example of great teaching..

[@ozyozk9466]

Hands down the best LQR tutorial I have ever seen.

[@magdiyouseff4661]

its the first time i understand the reason behind creation the objective function J. thank you for your clear explanation. thank you

[@christophberger599]

Really nice video, studying in vienna and did not get an idea how exactly the LQR worked, but now i understand. Thank you.

[@imanplus]

Thanks Dr. Abbott. That was the best explanation for LQR I've ever seen. Keep it up.

[@rezah336]

i just place all closed loop poles on the negative real axis to get no overshoot and adjust the speed by moving it along the axis. The speed is determined by the control signal.

[@cainghorn]

Speaking as a Msc in automation, I have never had someone explain this to me so clearly. Why would you ever go to a university, with pearls like these found on YT?:)

[@brucemurdock5358]

By far the best teacher for control systems and linear algebra (2nd one is arguable haha)

[@_electro_101]

Very well explained. I happened to see a general case of choosing R=1 and Q=C`C (c transpose*c). If so, Why?

[@alial-heji3282]

Thank you for posting this. Very easy to follow and understand.

[@luzltorai]

That is a great video. Thank Dr. Jake Abbott

[@giuseppealmontelalli840]

please make a video on LQG CONTROL

[@ericwindhede5937]

Thank you. Very nice way to introduce LQR. I´m looking forward to see you talking about Kalman Filter next. :)

[@salmann94]

One of the best explanations for the topic. Thank you!

[@guangweiwang9228]

Great physical intuition, thank you.

[@JadtheProdigy]

From a high level point of view, you give the LQR controller a trajectory X from 0:T, and a cost for trajectory deviation Q, a cost for effort R, and it returns the effort U from 0:T, as well as Xnew 0:T, such that the dynamic constraints are met, while J is minimized? In other words, X and Xnew may be different?

[@sumitshrestha2653]

Thank you! Sir. Can you upload a lecture on Parameter estimation: Batch least square, constrained least square and sequential least square?

[@vksateesh]

Very good explanation sir. Thank you.

[@georgevarghese5887]

This was vry well explained.. i need to know how can we use abc algorithm to decide the Q and R parameters..

[@HarishKumar-gw8bz]

Well, you can use the pso algorithm, which is quite easy to apply compared to the abc one.

[@devkumanan1773]

Great video, very well explained and easy to follow ! Thank you.

[@EricLikesTurtles]

Thanks for posting this. Very clear and helpful. Question: How do I select Q when one or more states is not observable, but is stable? For example, suppose the speed of a DC motor (generalized as a stable, first-order system with known dynamics and a voltage input) is itself the control input to a fully controllable, linear system. Furthermore, suppose that for whatever reason, the motor speed is not observable. When I append the the motor dynamics to the system, I now have an additional stable but unobservable state. Since I cannot measure the speed of the motor, I ultimately should have a feedback gain of zero that corresponds to motor speed. I assume that for the Q matrix, I would set the q term corresponding to the motor speed to zero. Is this correct? Thanks for your time.

[@25aaditya]

This is so amazing. thank you so much for a clear explanation of LQR

[@giancarlokuosmanen9723]

Awesome lecture, thanks!

[@kaizhou7331]

Very good explanation to a starter

[@mohamadmawed6078]

Your explanation is really amazing . Thanks a lot sir .

[@thibaultcantou3508]

Thanks a lot for those very clear explanations !

[@mohamadmawed6078]

Amazing and great explanation sir

[@dedenmusa]

Thanks Dr. Abbott. Greeting from Indonesia.

[@psjacome]

Thanks a lot for your tutorial. In the lecture you said it is necessary to simulate the system in order to choose Q and R values, Do you have some examples of simulations in Matlab??

[@jalpeshlimbola3958]

have you got the solution of choosing the values of Q and R?

[@p.z.8355]

So how would I change the control law to make it a tractor instead of regulator ?

[@PronoyBiswas]

Excellent video - very well explained

[@mohamedessamish]

a question plz if system is assym. stable and input excites a system then inputvanishes my question is will the states of sytem vanishes and what will happen to the output will it also vanishes ??? thanks for ur respond in advance

[@eatctitox]

Great explanation! I really appreciate you having a practical approach to explain the details... Super like!

[@AN-zk7kz]

Many thanks for the great video ! I would like to ask about the name of the method you followed to initialize the values of Q and R i.e. when you chose to start with the square of the maximum value inverse. Many thanks in advance !

[@antoniomdn]

Thank you very much for the tutorial, it explains it very clearly.

[@zyot511]

Thanks from Colombia. I really like your videos!

[@cendit420]

Thank you for the explanation, very helpful.

[@luisdamed]

Thank you!

[@klam77]

WOW ....clarity. i was also hoping for some intuition why they call it LQR! Basic math analysis says LQ is where you solve a quadratic ("Q") to meet a Linear ("L") constraint, but i don't see how that ties to Riccatti etc.....but otherwise you're video is TOPS!

[@klam77]

oop! I see it now. It's the same: a quadratic cost function (incorporating input and state) subject to linear (proportional to state) input! Linear-Quad.

[@bruno_sjc_]

You rock! What a clear explanation! Thank you very much!

[@kokorot17]

Thanks Sir.. you made my life a bit easier!

[@Pi314159265ify]

Great tutorial, thank you very much!

[@Sasipano]

thanks for a great introduction.... Moreover, I was curious of using it for my Inverted Pendulum Project.... So can you help me with that

[@bilalahmad-qo4wk]

I think that I can help you, what kind of problem are you facing?

[@alonealonesupervisor537]

+bilal ahmad brother do you have the textbook of this course

[@youssefabsi6296]

that was reaaally good. thanks

[@versatran01]

Nice video! Explains everything clearly! Great job!

Gave me help big time with my project topic, thumbs up!

[@bobanisback]

Thank you

[@BereketAbraham]

This video is amazing! Keep it up!

[@ahmadfariz7173]

Thank you, sir!

[@bhaskartushar90]

Great explained...

[@multimirage]

Well Said!

[@isaacsilva9631]

you are the best man. thanks very much from brazil.

[@viktorsawtschenko440]

Very nice video!

[@dmitriymakovkin]

Super! Clear & concise, thank you.

[@mohamadballout3847]

WOOOOWWWW!! Awesome!

[@lionconvoy8622]

very clear explanation! thanks! :)

[@Missionary117]

Wow.. that was very good.

[@wszolasss]

SOOO HELPFUL! THANK YOU!

[@pnachtwey]

+1 for saying zeros can cause overshoot even though the closed loop poles are on the negative real axis in the s domain. So now the problem changes from guessing where to put the closed loop poles to guessing on how to chose the weights. It seems like there needs to be yet another level of optimization to select the best weights for Q and R. It seems there are many possibilities for optimal or the term optimal is used loosely. Saturation isn't as big a problem as feedback resolution. BTW, it is possible to place zeros too.

[@ameerjanabi917]

awesome

[@Larantas]

Wonderful. Thank you :)

[@LuisRocha2]

Great tutorial, keep it up x)

[@spartanarmado]

Very usefull! thaks!

[@salmansircar5606]

Thanks your explanation was very helpful

[@zoro19840]

Really thankful .. (y)

[@LNasterio]

This is when you realize professors from some top end universities are JOKE!

[@manofsteeeeel-j1g]

+1

[@klam77]

well....they are just too comfortable with abstraction and leap ahead too quickly.....but to beginners it appears there are NO anchors in the material!

[@jalpeshlimbola3958]

theoretical understanding is ok and but the major problem about LQR method is that how to decide the perfect value of the matrix Q and R for to implement in the real physical system. if any one does know please let me know..thanks..!!

[@niteshagrawal486]

it depends on the system you are considering. Though main idea is to minimize or maximize J

[@brianblasius]

Use Bryson's rule. But this only gives you a starting point. So after using the rule, you can fine tune by hand.

[@NSAwatchesME]

well i understood none of that

[@avinashdamera6960]

thank you sir,