MPC support?

[roangel]

Hi guys, I see that you have added in the past some LTI, LTV and non-linear systems support (cartpole, floquet and neuralnet examples, respectively). I also see in the paper that you guys mention Differentiable-MPC, from Brandon Amos, and take it as a baseline. I have two questions:

• Does the current state of pypose already support MPC? I see it might not be the case because of this: Add model predictive control (MPC) #32
• Does pypose support hard constraints in the optimizations?

[wang-chen]

Hi @roangel Thank you for asking this, all the things you mentioned are currently under development, @aerogjy knows more about this.

[aerogjy]

Hi @roangel Thanks for the interest of our Differentiable MPC module! Currently, no. We will push the LQR module, which is the core part of D-MPC very soon, probably in a few weeks. Then, we will release the example we provided in the paper.
The constraint optimization is on our high priority development list. We are on-going developing it right now. Please continue keeping an eye on that.

[roangel]

Thanks a lot for the replies and clarification! I need the constrained optimization part, so I'll stay tuned. Is there any collaboration chat where I can stay posted and maybe contribute in the future? Thanks!

[wang-chen]

@roangel If you are pretty sure that you will contribute to the repo, we strongly encourage you to join our SLACK for discussion and task scheduling, otherwise you may also post any concerns, questions, and suggestions here, we will keep you updated.

[wang-chen]

@roangel MPC has been released in v0.5.0-beta.
You may install via

pip install pypose==0.5.0b0

It is a beta release because we will do coding acceleration later but the API will not be changed, so you may feel free to use it.
The forward function args u_lower, u_upper , du will be supported in a future release.

[roangel]

Thanks a lot for this! I'll take a look!

[roangel]

Hi guys, any idea when the support for du will be released? Thanks!