[Feature Request] Optional fall-recovery training for motion tracking policies

[XinyuSong123]

Hi,

Thanks for open-sourcing ProtoMotions. I am currently using it for humanoid / whole-body motion tracking, and I would like to ask whether it would be possible to support an optional fall-recovery training component on top of the existing tracking pipeline.

Recent works such as:

• Robust and Generalized Humanoid Motion Tracking
  https://arxiv.org/html/2601.23080v1

• A Kung Fu Athlete Bot That Can Do It All Day: Highly Dynamic, Balance-Challenging Motion Dataset and Autonomous Fall-Resilient Tracking
  https://arxiv.org/pdf/2602.13656

show that incorporating fall recovery into motion tracking training can improve robustness for highly dynamic, contact-rich, and long-horizon motion tracking.

Questions

1. Is fall recovery already supported somewhere in the current ProtoMotions codebase or configs?
2. Would this kind of optional recovery training feature fit the roadmap of ProtoMotions?
3. If not officially planned, could you point me to the most appropriate place in the codebase to prototype this?

Thanks again for the great work!

[tesslerc]

Yes this is something we will introduce very soon.
This is already implemented, just need to clean it up so it's fit for a proper release.

[zhang705]

Yes this is something we will introduce very soon. This is already implemented, just need to clean it up so it's fit for a proper release.

Looking forward to its official open-source release soon.

[bribrod]

Yes this is something we will introduce very soon.
This is already implemented, just need to clean it up so it's fit for a proper release.

Hello,

Thank you for your work on this system.

I'm using a MaskedMimic model to run characters in my video game. My characters keep falling down and cannot get up. #LifeAlert
Can you pease publish a trained model with the get up magic?

Thank you,
Brian