google-deepmind / mujoco_mpc

Real-time behaviour synthesis with MuJoCo, using Predictive Control
https://github.com/deepmind/mujoco_mpc
Apache License 2.0
1.06k stars 157 forks source link

Request: open Discussion forum on this repository? #320

Open MyelinsheathXD opened 6 months ago

MyelinsheathXD commented 6 months ago

Hello everybody! Hey, this is amazing tech and need to be have a conversational questions about the system rather than only issues on the repository.

It would be greate if there is a Discussion forum on this repository like mujoco repository. Could you enable Discussion forum on this repository?

thowell commented 6 months ago

@yuvaltassa

yuvaltassa commented 6 months ago

I tried, I swear. Couldn't figure it out 😛 Paging @nimrod-gileadi

shivanshpatel35 commented 2 weeks ago

I also feel that a discussion forum will be super helpful. The overall codebase is more complex than other simulation codebases, so a discussion forum will help reach this work to more people faster.

yuvaltassa commented 2 weeks ago

@nimrod-gileadi, @thowell, how do we do this?

ShahhhVihaan commented 2 weeks ago

Yup, I would love this too!

I understand that the repository currently supports shooting-based planners, which are ideal for real-time applications. However, I’m wondering if you would consider adding collocation methods as an option for users who need higher trajectory accuracy, even if it means sacrificing real time performance. If so, I would love to contribute.

I’m relatively new to optimal control and would like to implement both shooting and collocation methods (direct and indirect) from scratch, inspired by MJPC’s structure. I could fork MJPC and work on it independently or, if you’re interested, I’d be happy to contribute directly to this repo.

If this is something you’d like to discuss, we could explore it further in the discussion forum.