Data-Driven Control: ERA/OKID Example in Matlab

No video

Data-Driven Control: ERA/OKID Example in Matlab

Рет қаралды 8,600

6 жыл бұрын

In this lecture, we explore the observer Kalman filter identification (OKID) and eigensystem realization algorithm (ERA) in Matlab on an example.
Code: faculty.washington.edu/sbrunton/DataDrivenControl.zip
www.eigensteve.com/
This video was produced at the University of Washington

Пікірлер: 14

@theidealisticman 2 жыл бұрын

Prof. Brunton, you are a legend.

@Eigensteve 2 жыл бұрын

Thanks!

@BalajiSankar Жыл бұрын

Thank you for the video on using era and okid

@luiggitello8546 8 ай бұрын

So much satisfaction to reach this point

@leoliu3847 3 жыл бұрын

Thank you for the great video. Question for you, Dr. Steve? Would you please explain how you get the impulse generated for a MIMO system seems like you are doing a DOE of Nu by Nu and combine them instead of doing that in one time?

@hrt7180828 2 жыл бұрын

Dear Steve, in the ERA function you used Ur to calculate Br. Also, you used Vr to calculate Cr. These should be replaced

@mateogonzalezhurtado1094 4 жыл бұрын

What if I only have the output data? How can I identify the system without the input data?

@tayyipensarozkaya2329 Жыл бұрын

I am still waiting for someone to email me my dataset with a randomly forced u.

@hrt7180828 2 жыл бұрын

a fascinating result is that your code despite all these replacing and the power of sigma which should be 0.5 but it is -0.5, works properly and when I change it to the correct version it does not! I also check with your book, the same thing you mentioned in the video is also in the book. So, can we conclude that what you explained in the book and in the video are incorrect and what you wrote as the code is correct?

@markghali7623 4 жыл бұрын

Could you please update the code on the course website ?

@benjomino87 2 жыл бұрын

Is the matlab code available to the public? where can I get it?

@markghali7623 4 жыл бұрын

How does one decide the order of the resulting system ? through SVD ?

@sophiakim107 4 жыл бұрын

Hey Mark - I think you're right, but more specifically, it depends on how well you want your data to estimate your full state model. In lecture 10, Prof. B talks about how you can look at the architecture of your singular values. If your singular values suddenly drop off at one point, then you know that that will probably be a good place to truncate. If you have a desired value for error, I believe you can extract r from the upper and lower bound of the infinite norm of G. I also think lecture 13 shows a good example how how much information your first r modes contain. At 2:00 I think the cumulative sum plot shows how much of the Hankel singular value energy is encapsulated in the first r modes.