Programmatically generating Simulink models from symbolic equations

For control design and analysis, I am a big fan of the connect() function in the Control System toolbox. It's quite convenient for building up a large linear dynamic system without having to resort to lots of manual linear algebra (or use of the lft function). I simply define my blocks and the InputName, OutputName properties and MATLAB takes care of the rest.

I would like to do something similar for non-linear dynamical systems. To my knowledge, MATLAB doesn't offer this sort of functionality. So I wrote some primitive scripts that allow me to define non-linear dynamic systems and interconnect them just using symbolic substitution. The resulting blocks have the following fields (example 3 state system shown)

       .StateName: {'iL_F2'  'vC_F2'  'vCd_F2'}
       .f: [3×1 sym]
       .g: [2×1 sym]
       .InputName: {'v3'  'i4'}
       .OutputName: {'i3'  'v4'}

The equations f and g contain the differential equations and output equations respectively (i.e. d/dt(x) = f(x,u), y = h(x,u)).

These equations are symbolic, but I would like to simulate them in Simulink. Currently I manually paste the differential equations in a block, define my inputs and outputs to make the .InputName, .OutputName properties, and then hook up an integrator to integrate d/dt(x) = f(x,u).

I'd like to do this automatically in code but I'm clueless as to how to populate a model with user-defined differential equations. To be clear, I know the basics of building a Simulink model programmatically and setting parameters. But I'm unclear how to 1) turn my symbolic equations into the right format and 2) get them into a user-defined block.

Any thoughts on this would be appreciated. Even better, any links to existing methods for automatically generating Simulink blocks from non-linear differential equations. I think someone must have already developed a toolbox for doing this, right ?

Answer 1

[Nicolas Schmit]

1. Convert the symbolic functions to MATLAB functions and generate .m files using the matlabFunction command.
2. Call the .m files from a MATLAB Function block in Simulink.

[jdg]

Thank you for the quick response. This seems to be exactly what I need.

Answer 2

[Tim Grunert]

In the last year we also the problem that we need to model twice: symbolically for control analysis and in simulink for simulation. To make it more easy we developed a small tool solving this problem. It can be downloaded as free software under: https://github.com/odesca/

Answer 3

[Walter Roberson]

You may wish to look at odeFunction() along with the documentation where it shows calculating mass matrices and so on.

[jdg]

Thanks! Surprised I never learned of these tools before.

[Walter Roberson]

I only learned about them a few months ago myself. The steps seem overly complicated sometimes, but they do work.