Adding inputs and outputs functions #1209
Conversation
src/systems/abstractsystem.jl
Outdated
| end | ||
|
|
||
| function outputs(sys::AbstractSystem) | ||
| [st for st in states(sys) if isoutput(st)] |
There was a problem hiding this comment.
This may not be true. Do you want it to be true?
There was a problem hiding this comment.
In what case would it operate incorrectly?
There was a problem hiding this comment.
yeah they could be eliminated to observed?
There was a problem hiding this comment.
ah, so should I cat any states in observed that are also decorated as outputs?
There was a problem hiding this comment.
But then, you cannot calculate the Jacobian.
There was a problem hiding this comment.
Suppose we have the reduced output as an observed variable x ~ a + b with the equation D(y) ~ a + sin(a) + b. What's the derivative of the right hand wrt x ~ a + b? We cannot differentiate this thing. First, matching a+b is hard. Second, we don't know if this a + b is from x. It could very well be something else.
There was a problem hiding this comment.
I disagree with you on the idea that the a+b needs to be "from" x, but I see your point nonetheless. Does that have anything to do with it being an output though? Your argument seems to be valid regardless of that. Right now, don't we just expect people to, for lack of better words, "not do that"?
There was a problem hiding this comment.
Given x is an output, what's the output Jacobian of
D(y) ~ a + b + x
0 ~ a + b - xand what's the output Jacobian of
```julia
D(y) ~ 2a + 2b
wrt a+b?
There was a problem hiding this comment.
Say x isn't an output though. Isn't this still a problem? I'm failing to see where it being labelled as output=true comes in
There was a problem hiding this comment.
When linearizing control problems, we need to compute the Jacobian wrt the inputs or the outputs.
This adds functions to access inputs and outputs of a system