Optimal Control API about modelingtoolkit.jl HOT 14 CLOSED

Yup thanks for the writeup. So it looks like a two variable + parameter pairing is what's needed. DependentVariable for states and Variable for the control variables would work, and we can just add the aliasing between the names DependentVariable and StateVariable along with an alias of ControlVariable and Variable.

from modelingtoolkit.jl.

Actually we'd want to keep Variable separate for intermediate calculations, so we should add ControlVariable and that would make it easy to automatically detect that it's an optimal control problem.

from modelingtoolkit.jl.

Yeah, that sounds good. Also, ultimately it is important for me to convert whatever the user gives into expressions so that I can use JuMP, is that out of scope for this package?

from modelingtoolkit.jl.

That's in scope.

from modelingtoolkit.jl.

@ChrisRackauckas great!

from modelingtoolkit.jl.

Also, it would be good to extend your @Var macro to define new names for the state and control variables such as:

@StateVar X Y PSI

and

@ControlVar U AX

I did this before using a function

from modelingtoolkit.jl.

Is @DVar fine? It would make interopability easier with DiffEq to not add another name if it's unnecessary.

from modelingtoolkit.jl.

There should be a difference between state and control variables. Are you suggesting using @Dvar for the states and controls?

from modelingtoolkit.jl.

No, just states, since those are just the standard diffeq variables?

from modelingtoolkit.jl.

yeah, that is fine! What does @Dvar stand for? It sounds like design varaible

from modelingtoolkit.jl.

Dependent variable.

from modelingtoolkit.jl.

ah, that works!

from modelingtoolkit.jl.

Could @Var be used for the controls?

from modelingtoolkit.jl.

We'll want that for intermediate calculations.

from modelingtoolkit.jl.