-
|
Hello Theseus, First, congrats for the amazing job! This library is very promising. I wanted to ask a simple question, I read the paper and it says that you don't support equality constraints in your optimization problem layers. Does this mean as well that users can't enforce dynamics (x_{k+1} = f(x_k, u_k)) in the optimization problem? Thanks! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
Hi @roangel, thanks for your interest. Yes you are correct, at the moment we only have support for unconstrained nonlinear least squares objectives. In practice typically we can just apply dynamics as a weighted soft constraint, || w (x_{k+1} - f(x_k, u_k)) ||^2 similar to these cost functions and are used in the motion planning example. Support for constrained solvers is planned for the future. |
Beta Was this translation helpful? Give feedback.
Hi @roangel, thanks for your interest.
Yes you are correct, at the moment we only have support for unconstrained nonlinear least squares objectives. In practice typically we can just apply dynamics as a weighted soft constraint, || w (x_{k+1} - f(x_k, u_k)) ||^2 similar to these cost functions and are used in the motion planning example.
Support for constrained solvers is planned for the future.