Remove dead up-axis conversion from MuJoCo solver

[adenzler-nvidia]

Summary

• Removes leftover Y-up→Z-up coordinate swizzle from update_body_mass_ipos_kernel
• The MuJoCo solver spec is built entirely in Newton's native coordinate system (body transforms, joint frames, gravity all passed through as-is), so converting the COM position to Z-up was incorrect — it put body_ipos in a different coordinate system from everything else
• Also removes the matching incorrect test expectations and an unused shape_bodies variable

Closes #229 — the scattered pos2mjc/rot_y2z conversions referenced in the issue have all been removed. The remaining up_axis usage across the codebase (importers, geometry SDF kernels, viewer, mesh utils) properly handles all 3 axes (X, Y, Z).

Test plan

• All 141 MuJoCo solver tests pass (uv run --extra dev -m newton.tests --no-cache-clear -k test_mujoco_solver)
• CI passes

Summary by CodeRabbit

• Refactor

  • Streamlined inertial property calculations by simplifying coordinate transformation handling in the MuJoCo solver.

• Tests

  • Updated test cases to reflect simplified calculation approach.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

---

📝 Walkthrough

Walkthrough

This PR removes the up_axis parameter from the update_body_mass_ipos_kernel function signature and eliminates conditional coordinate reorientation logic. The corresponding kernel caller in SolverMuJoCo no longer passes the up_axis argument, and test expectations are simplified to use untransformed Newton model coordinates directly.

Changes

Cohort / File(s)	Summary
MuJoCo up_axis parameter removal newton/_src/solvers/mujoco/kernels.py, newton/_src/solvers/mujoco/solver_mujoco.py	Removed up_axis parameter from update_body_mass_ipos_kernel signature; eliminated conditional branch reorienting COM coordinates based on up_axis; body_ipos now always set directly from body_com without axis-based transformation.
Test coordinate conversion cleanup newton/tests/test_mujoco_solver.py	Removed up_axis-based position and orientation conversions; simplified test comparisons to use untransformed Newton model values (newton_pos, newton_com) without Y-up/Z-up reordering adjustments.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• PR #779: Directly related—both PRs modify MuJoCo solver inertial/axis handling by removing/upstreaming up_axis-based reorientation in kernels and solver calls.
• PR #1199: Related—both PRs modify the update_body_mass_ipos_kernel signature and its parameter handling.

Suggested reviewers

• eric-heiden
• jvonmuralt

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Linked Issues check	❓ Inconclusive	The PR removes specific up-axis conversion code but does not implement the broader refactoring proposed in #229 to support X-up, Y-up, Z-up consistently or introduce shared functions.	Clarify whether this PR fully closes #229 or is a prerequisite step; if partial, consider opening a follow-up issue for the shared function refactoring mentioned in #229.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: removing up-axis conversion logic from the MuJoCo solver kernel, which is the primary focus of the changeset.
Out of Scope Changes check	✅ Passed	All changes are directly related to removing up-axis conversion from the MuJoCo solver, with no unrelated modifications detected.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[adenzler-nvidia]

I'm not very confident on this one because I don't really remember the full up-axis story. But the code looked inconsistent at least, because we were only rotating ipos.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!