How to design MPC with custom estimator and optimizer

Please I need opinions on how to

1. design MPC controller for a plant using a custom optimization technique and custom internal state estimator. I am thinking of using neural network to estsimate the internal plant state.

2. Can I also use the MPC block in the simulink library without using the in-built state estimator and optimizer

Answer 1

[Sam Chak]

Hi @Kamal

If you can design a generic state feedback controller for a plant using a custom optimization technique and a custom state estimator, I believe you can apply the solution approach to the custom MPC design process. Consequently, you are free to create the MPC algorithm based on your personal, unconstrained understanding without using the built-in mpc() function or the MPC Controller block in Simulink.

I assume that you are proficient in certain optimization techniques and state estimation. If there is no novelty involved, could you please provide a sample MATLAB code to design the state feedback controller for a simple, undisturbed Double Integrator system ?

[Kamal]

Thank you for the answer. Actually I am just thinking of an approach to introduce novelty into an MPC design for a system since MPC design has been executed by several researchers and designers. I am a beginer in MATLAB.

Thank you so much

[Kamal]

The oprtimization I had wanted to apply is called NEORL. Neuro optimization with reinforcement learning and I thought of deep nueral network for estimating the system dynamics. Howver, I have decided not to continue again.