Seeking MuJoCo feedback on the new OpenUSD standard for robotics (REP-158)

[adamdbrw]

Hello Everyone,

A new standard proposal has just been released, and I would love to engage the MuJoCo community in the discussion.

The standard is OpenUSD Conventions for Simulation Asset Interoperability in Open Source Robotics. The goal is to ensure a single 3D asset works seamlessly across different physics engines (like Gazebo, O3DE, Isaac Sim, Newton, MuJoCo, and Genesis), ROS runtimes, and web visualizers.

While MuJoCo relies natively on MJCF and works with OpenUSD primarily through conversions, this proposal is highly relevant for our ecosystem. It focuses on:

1. Ensuring reliable import/export paths with minimal data loss.
2. Standardizing "Simulation Ready" practices across different engines.
3. Handling physics semantics in an engine-neutral way.

Since MuJoCo has unique physics features and its own USD export tooling, feedback from this community will be incredibly valuable to make sure the standard works for everyone. I invite you to join the discussion here: https://discourse.openrobotics.org/t/draft-rep-158-openusd-conventions-for-simulation-asset-interoperability-in-open-source-robotics/55526.

[havess]

Hey Adam! I'm excited about your proposal, particularly to bring ROS concepts under the USD umbrella seems like a great step forward thanks for getting the conversation going. I do want to elaborate a bit on MuJoCo's OpenUSD stance.

MuJoCo has it's own set of USD schema that map very closely to what MuJoCo exposes to the user via MJCF. This was intended as a first pass at USD support in MuJoCo beyond the current limitations of UsdPhysics.

We have been actively engaging with the Newton USD Schema and consider these a staging ground for more general solver agnostic description. It's my opinion that due to the nature of Newton, if a schema serves it's purpose there we can assume enough generality to push some aspects upstream to UsdPhysics. As Newton adopts more schema, our plan is to deprecate any redundancies in mjcPhysics and the same will apply with any upstream changes to UsdPhysics.

Regarding ext_physics I would however like to exert a bit of caution that we do not duplicate work and end up with some assets that use the "neutral" Newton schema and some that use this "neutral" ext_physics schema. I'd also like to put out a call to action, if there are bits of the Newton schema that do not suit your needs, or if you think we have made a mistake please do open an issue. We have the flexibility to move fast with these schema, and are happy to receive input from the community.

[adamdbrw]

Good points there, Sam. Ext_physics seems the most controversial (and rightly so) part, but in the REP submission, it is only auxiliary. I will gladly adopt the namespacing and the extent to recommendations from both AOUSD and simulators communities.

I would love this REP to map very well to a forward-looking take on OpenUSD that MuJoCo already embraces, and to make things even better for the community through cross-simulation standardization.