Open dfridovi opened 3 years ago
It's just a question of which unit you sign to u
. I thought of the input as u = [Δθ, Δv]
. Thus, they have the same unit as the state dimensions and no time scaling should be necessary. Am I missing something?
I think it’s technically ok - it’s just scaling u by something that’s constant in each problem. The standard model would be like x dot = f(x, u) so u is automatically in units of state dot if there’s a state who’s derivative is u.
Sent from my iPhone
On Mar 5, 2021, at 3:07 AM, Lasse Peters notifications@github.com wrote:
It's just a question of which unit you sign to u. I thought of the input as u = [\Delta \theta, \Delta v]. Thus, they have the same unit as the state dimensions and no time scaling should be necessary. Am I missing something?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
I see, this perspective makes sense if you view the problem as a discretization of the continuous time dynamics. We can certainly change that if it improves clarity.
Yeah I would do that just for consistency with standard notation. It doesn’t change the results at all - like you just need to scale u costs equivalently to make it the same.
Sent from my iPhone
On Mar 5, 2021, at 8:50 AM, Lasse Peters notifications@github.com wrote:
I see, this perspective makes sense if you view the problem as a discretization of the continuous time dynamics. We can certainly change that if it improves clarity.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
https://github.com/lassepe/JuMPOptimalControl.jl/blob/18cf096459f912358f6d1eabb3c65f4fccd83488/test/utils/unicycle.jl#L42
Is
\Delta T
missing here? I'm pretty sure this will just have the effect of needingu
to be larger...