Model Predictive Control Design Parameters | Understanding MPC, Part 3

[@MATLAB]

Download this hands-on MPC virtual lab to practice design of model predictive controllers for an autonomous vehicle steering system: bit.ly/MPC-virtual-lab

[@seraph_v]

This is the best control video that I have ever seen in my life

[@MATLAB]

That's awesome. Thank you for watching!

[@abhinav_mzn]

This is the best control video that I have ever seen...

[@fernando.liozzi.41878]

Excellent explanation! Could you please explain with a practical example using Simulink? Thank you very much.

[@abreu9999]

Great explanation and an excellent teacher. Thank you.

[@surajveertalreja9438]

Excellent Lecture with Excellent Examples. Thank You so much!

[@tjdwnd9529]

Fantastic. now I understand the concept of MPC and it took only 30minutes

[@hyunkiseong1445]

Nice video to understand the MPC! I also hope to learn with practical examples of Simulink. Thank you so much.

[@davidkooi4349]

Fantastic!
Thank you Melda and MathWorks

[@joht91]

Nice illustration and very good explanation. However, I don't really follow the part about control horizon. If the setpoint varied(instead of being constant as described in the video) for each time instant k, a control horizon shorter than prediction horizon is not suitable, right?

[@meldaulusoy8389]

Hi, unless you have preview, meaning your controller knows about the future reference, the setpoint used in the prediction is assumed to be constant. The controller looks at the value of the (varying) setpoint at current time step and uses it throughout the prediction. So, the same discussion in the video about control horizon would apply. You wouldn't choose a large control horizon because this increases complexity and doesn't significantly improve controller performance.

[@joht91]

Hi Melda,
Thanks for the answer.

[@goodwords.5645]

please do a video on adaptive mpc for dc dc converters

[@tunabilgin1993]

Great explanation, though I am confused about one aspect. Don't we throw away our previous computations at each time step regardless of any disturbances that might happen?

[@francismikaelmagueflor1749]

yeah you throw them away but you're wasting computational resources by having an even larger window when those predictions can be easily made void by disturbances

[@권세빈-u4v]

thank you for video

[@金飘飘-o2o]

amazing video!!!!!

[@alexandrpetrov1110]

now I understand the concept of MPC and it took only 30minutes

[@kumarrahulbhadani]

I am eagerly waiting for examples using Model Predictive Control Toolbox in Simulink.

[@meldaulusoy8389]

Hi Rahul, we will post a couple of more videos, two of them focusing on different flavors of MPC such as adaptive, gain-scheduled and nonlinear, and methods to run MPC faster. And after these, we'll have Simulink demo videos. Thank you for you patience!

[@thugonomicsstark7238]

if my reference is coordinates (x,y) and my plant (mobile robot ) has (x,y,theta) how do i implement it , please answer!

[@benjaminaskazlauskas1080]

Thank you

[@mohemahmoude3356]

am a bit confused,does the sample time equal to prediction horizon

[@meldaulusoy8389]

They are different. The sample time of the controller is really about how long it takes for the controller to compute an output to the plant. And it depends on several factors such as the system bandwidth and the maximum sampling rate of your processor. Prediction horizon is measured either as time duration [seconds] or time steps as in the video. It is about how many sample times you predict into the future.

[@Manish-fo4gj]

Good explanation

[@kumarrahulbhadani]

And please use examples from Autonomous drive while demonstrating MPC toolboxes. :)

[@hamztoufik9984]

thank you for the explanation and sincerely you are very beautiful .

[@asifshehzad9921]

mam you are too beautiful

[@av6966]

Get a grip! It's a not a disco!