L3.3 Introduction to Model Predictive Control (MPC) - regulation

[@marofe]

this is the best explanation of MPC control! thanks

[@jakedesmond5107]

Great video series, thanks a lot. I did have one question however. Around 2:48, when you moved one time step into the future and updated the current state of your model, you assumed that the model prediction was exactly the measured response of the plant's state. I would assume that this is rarely the case, i.e. the measured state is not exactly what the model predicted. How does MPC account for these disturbances (e.g. the drag affecting a car's acceleration is different from the model's prediction and thus you aren't applying the correct amount of throttle for a desired acceleration). I see in your next video you discuss reference tracking, but it seems like the outputs are just the model's predictions and not the actual measured states. When (if ever) does the model compare the signal is commanded versus the difference in the model's predicted response versus the measured response? Thanks.

[@aa4cc]

Jake, your observation is perfect. Indeed, my explanation at that part of the video seems to suggest that at the next time step the optimization considers as the initial state for the shifted time window the value predicted from the model in the previous step. I should have somehow reflected in the video that this is actually not the case because the state at the next/new time step is either measured or estimated, thanks to which MPC becomes effectively a feedback control. Thanks for pointing this out, it might really be confusing. I will perhaps incorporate this explanation in the next version of the video. The fact also is, however, that similarly as in the case of other feedback control strategies, the actual (measured, estimated) state (affected by some disturbance) should not deviate from the predicted (using a model) too much. If the disturbance is severe, some sort of disturbance estimator should be implemented.

[@bdasama7893]

Can you give me document explique all that ?

[@Carlotta639]

Great lecture, thank you so much.

[@ukulele2345]

best explanation on youtube!

[@NKPyo]

you missed 6 in the chart you drew.

[@aa4cc]

Indeed. It is funny:-) Well, these mistakes happen to me - apparently the mind was already focused on something else while labelling the axes - but I hope these typos/mistakes do not hinder the understanding.

[@zPauLoma]

Thank you so much. Really good explanation =).