THANK you so much for your content here on youtube. Were it not for you i would still have so many reexams to take. This means that i can end my master's thesis this summer. Cant wait!
Just like how the linearization of a non-linear system might cause a control system to fail outside certain bounds...do estimators also fail in certain situations? What happens in that situation?
Hi sir, Could you give me an example for noise, disturbance in real system? What did make them? For example, to control speed of DC motor, what will be disturbance, noise? Thank you in advance!
im not sure, but i guess there would be emf noise on your sensor (depending on what sensor you use), if you are using an IMU, there might also be vibrations from the motor which can vibrate the IMU and therefore introduce some error in your sensor readings. You might even get bad readings from your sensor as well. Also, since this is not a perfect world, you can always expect that your measured values will never be 100% accurate and the readings themselves are a kind of "best guess" of the sensor.
Of you're using a high cpr encoder, the position graph won't be smooth, it will be very erratic. Using accelerometers to get position is a good example of noise, because of the imperfections the reading will never be 100% accurate, and integrating twice to get position will just accumulate all the errors leading to wildly inaccurate position estimations given enough time.
Hİ Professor, I observed that if disturbance noise is much smaller than a certain level of observation noise, the estimation performance is much worser than the scenario that for the same level of observation noise, disturbance noise is close to the observation noise. I don't think this is a wrong observation since in the first scenario, we trust on our model and observations are not much important, so the estimation performance is bad but for the second scenario, we make significant corrections based on our observations, but how can this be shown mathematically?
I think it depends on what you want to see. some readings may be way off the real value so I don’t think you can use mathematics to tell you what values what’s the most convenient filter…
best videos in modern control engineering on YT
Thanks for sharing Prof. Finally found the clearest video in explaining kalman filter
Glad it was helpful!
THANK you so much for your content here on youtube. Were it not for you i would still have so many reexams to take. This means that i can end my master's thesis this summer. Cant wait!
from the very first seconds of this video, I knew it's gonna be a good one !
Thanks!
I'm not even graduated but I love to learn with you! Kalman Filter was one of my biggest fears by far🤪
where is the "welcome back"
=) everyone is always welcome!
Just like how the linearization of a non-linear system might cause a control system to fail outside certain bounds...do estimators also fail in certain situations? What happens in that situation?
That noise you make is unbelievable
one question, how do you write from right to left??
Years of practice :)
After the video is finished, apply 'mirror'. But, you have to be a lefty.
its thing called lightboard :]
His wedding ring is on the wrong side, that tells you it's flipped in post
this is what exactly I was looking for!!!
You are a legend
Hi sir, Could you give me an example for noise, disturbance in real system? What did make them? For example, to control speed of DC motor, what will be disturbance, noise? Thank you in advance!
im not sure, but i guess there would be emf noise on your sensor (depending on what sensor you use), if you are using an IMU, there might also be vibrations from the motor which can vibrate the IMU and therefore introduce some error in your sensor readings. You might even get bad readings from your sensor as well. Also, since this is not a perfect world, you can always expect that your measured values will never be 100% accurate and the readings themselves are a kind of "best guess" of the sensor.
Of you're using a high cpr encoder, the position graph won't be smooth, it will be very erratic. Using accelerometers to get position is a good example of noise, because of the imperfections the reading will never be 100% accurate, and integrating twice to get position will just accumulate all the errors leading to wildly inaccurate position estimations given enough time.
can you please have a full playlist on Kalman filters?
Very nice introduction, great intuition. Thank you.
This makes me realize how bad my professor really is, thanks for sharing.
I need solution of kalman-yakubovich conjugate matrix equation which can be expressed as polynomial of matrices.
This was crisp and clear. Thanks a lot
God damn this course is on point 👌
Hİ Professor, I observed that if disturbance noise is much smaller than a certain level of observation noise, the estimation performance is much worser than the scenario that for the same level of observation noise, disturbance noise is close to the observation noise. I don't think this is a wrong observation since in the first scenario, we trust on our model and observations are not much important, so the estimation performance is bad but for the second scenario, we make significant corrections based on our observations, but how can this be shown mathematically?
I think it depends on what you want to see. some readings may be way off the real value so I don’t think you can use mathematics to tell you what values what’s the most convenient filter…
I was shocked at how you were writing backwards until I saw your ring
😊😊😊
Thank you very much )))
The best
Thank you too!
Thanks, Professor!
My pleasure!
Who knew KFC was something so complex :P
“Gow-see-ann”