Get nonlinear model from Simulink to Matlab

Is there a way to get my nonlinear Simulink into a Matlab object that I can than use to create a nonlinear MPC? The model is additionally time variant, but it would already help me to just get a nonlinear model in Matlab.

I only found the way of either creating linearized state space models for multiple snapshots that represent the time varying aspect good enough, or alternatively describe my whole model in Matlab directly and create a nlmpc object.

Answer 1

[Emmanouil Tzorakoleftherakis]

Hi Daniel,

Unfortunately Model Predictive Control Toolbox does not currently support using a Simulink model directly as a prediction model in MPC. However, as a workaround, you can train a data-driven model with neural ODEs to capture the dynamics that you need, and use this data-driven model for prediction. This functionality requires System Identification Toolbox and Deep Learning Toolbox.

Here is a video and some code that shows how this works for a simple pendulum system