lunduniversity.lu.se

# Predictive Control FAQ

Q: What questions are collected here:

A: A question will be considered "frequent" when it has been asked once. If it is a good question. I will anonymize the questions, and possibly perturbe them if that surves my purpose :-)

Q: Will you put up the lecture slides in advance?

A: I will try to. But my guess is that it might be quite late I do it. I keep fine tuning most lectures until late.

Q: What if I feel somewhat uncertain about the mathematics for discrete time systems etc. Is there some material I can read, besides appendix A,B in the book, to get in pace with the other students?

A: If you take the real-time systems course in parallel, you will find some material in the beginning of that compendium (Computer Control: An Overview Educational Version 2012, buy at the KFS campus bookstore). You easily also find some online resources if you look around, such as this.

Q When will the lab signup lists be available?

lab1 is available from Sep 11. We will aim to have the signup lists available a week before the labs start.

Q: Hej, (English translation: "Is there a typo in the homework1 problem")

Har börjat kika på den första hemuppgiften och har en fråga kring uppgift två. I uppgiften står det H_2(z) = 1 / (z + 2z + 1), att separera z från 2z får mig att misstänka att det borde ha vart H_2(z) = 1 / (z² + 2z + 1). Det jag undrar är därför om det är ett tryckfel eller om det bara är jag som blir förvirrad över sättet det är skrivet på? (

A: Oops. Yes you are correct. It should be z^2. I have updated the homework text now.

Q: In Homework 1 it says one should use matlab to sample the system. How do I do that, I dont see how I can integrate matrices in matlab.

A: The easiest is to use the command "c2d".The answer should be numerically the same as what you get from your hand calculations done before.

Q: What is the difference between "q" and "z"? Are they the same?

A: It's the same difference as between the d/dt-operator and s-operator in continuous time: If you have dy/dt this becomes sY(s). The "q" operator gives a forward time shift of a time series, e.g. qy(k) = y(k+1). The "z" operator gives a similar shift, but on the z-transform of the signal instead. When you z-transform, all q are changed to z. If for example y(k+1)+ay(k)=u(k), this can be written either as (q+a)y(k)=u(k) or as (z+a)Y(z) = U(z).

Q: Where can I read about  MIT-rule, and what does it stand for? I havent seen in the lectures or the book yet. It is mentioned in exercise 4.

A: We mentioned the MIT quickly on Exercise 1, but probably not enough:

If you have an error e, such as e = y-y_m, which depends on some parameter(s) then the MIT rule is to move in the negative gradient of J(e)=e^2, which becomes

dtheta / dt = - gamma * e * de/dtheta

I should have had a slide about it on the lectures. Sorry. For those of you that has bought the extra book, Adaptive Control, it is mentioned there. The name comes from the MIT affiliation of the inventors

Q: Is there an error in exam 2007 6c?

A: Yes. It says | 1-k exp(iwh) -0.5| < 1 but this should be | 1-k (exp(iwh) -0.5)| < 1.
Unfortunately this means that there is no such suitable k-value. (Compare alsowith exam 2009 problem 6.)

Page Manager: