Problem with Model Predictive Controller Design

Hello, I am new to controller design. For my research work I am developing discrete time linear parameter varying state space model based model predictive controller. I have a linear parameter varying state space model in discrete format. My first question is If i want to simulate the controller, Do I necessarily have to have a simulation plant (apart from State Space) or real sensor outputs ?

My second question is does Adaptive MPC always takes (A,B,C,D,U,Y,X and DX) as input because I have (X and Xk+1) in my discrete state space format?

Answer 1

[Rijuta]

Hi Mohit,

I understand to want to know if you need a simulation plant to simulate a controller and also understand about Adaptive MPC.

To simulate your model predictive controller in MATLAB, you will need a simulation plant that is consistent with your linear parameter varying state space model, and you can incorporate measurement noise to simulate sensor outputs. If your linear parameter varying state space model includes the state vectors x(k) and x(k+1), you can use these vectors as inputs to the MATLAB function that creates the Adaptive MPC controller. The specific function you use will depend on the MATLAB toolbox you are working with.

Please refer the documentation to know more about Adaptive MPC.

I hope the above explanation answers your question.

[Mohit Punasiya]

Hii thanks for the answer,

I understood the first part,

For second part, I don't think I made my self clear before. let me try again - The AMPC block takes dx as one of the inputs while the model and controller is in discrete setting, what I did is feeding dx = X(k+1)-X(k) , is it correct?