examples/mujoco_xr: pick the gripper ghost with --robot

[jiwenc-nv]

Description

Stacked on #900 — review that first; the base here is jiwenc/mujoco-xr-scene, and #928 is a sibling on the same base (it changes _update_ghost's signature, so one of the two will need a trivial merge).

Adds the reBot DevArm gripper beside the SO-101 leader gripper, selected with --robot={so101,rebot}. Everything that differs between them — scene, meshes, mocap bodies, how each moving part is driven, where it sits on the hand — moves into a new robots.py, so app.py keeps only the machinery and a moving part is a hinge or a slide. Those two cases are the whole of _update_ghost.

The two ghosts are not the same kind of thing, and the placement follows. The SO-101 leader is the device the fist is closed around, so its posture stays a headset measurement with no test pinning it. The reBot leader is a back-driven arm on a table, so what the controller commands is the follower's parallel jaw — nobody holds it, there is nothing to tune, and its placement is the OpenXR grip frame's own definition (ghost +x → grip −Z, the pen-tip axis; ghost ±y → grip ±X, so the jaws open the way the index finger squeezes) with the fist on the drive motor. A test may pin that, and does. This is the call most worth challenging.

Two things came out of the geometry rather than a preference. Upstream disagrees with itself on the jaw travel — joint_left says 0.05 and joint_right 0.0715 for one rack-driven pair — and the cnc7 rail plate breaks the tie: at 0.05 the carriage's outer edge stops 3.4 mm inside the plate's end, at 0.0715 it hangs 18.1 mm past it. And the jaw origins are written as quaternions rather than copied from rpy, because URDF rpy is fixed-axis Rz·Ry·Rx while MuJoCo euler= is intrinsic Rx·Ry·Rz, and these have two non-zero angles. Both are checked against the fetched URDF.

Licence wrinkle worth a decision. Meshes come from Seeed's own reBotArm_control_py at a pinned, checksummed commit — the only source with both the gripper geometry and the URDF the tests read. That repository declares MIT in its README.md and ships no LICENSE file, so the fetch script pulls the README in its place. The hardware itself is Seeed-Projects/reBot-DevArm, CERN-OHL-W-2.0. Say the word if you would rather it came from somewhere else — it is one pinned line in a 70-line script.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update

Testing

Ubuntu 22.04 / aarch64 (Tegra). Full cmake --preset py3.12 -DBUILD_VIZ=ON build green; ctest -L mujoco_xr green — 4 files, 57 tests, all executed rather than skipped. The generic ghost assertions are parametrized over both robots; the reBot ones cover the URDF-derived jaw frames, the travel, the closing sweep and the placement convention. uv pip install ./examples/mujoco_xr was checked to carry both asset directories and the extension into the wheel.

The rail-plate assertion was checked to actually discriminate: it reads −3.39 mm at 0.05 and +18.11 mm at 0.0715, and fails at the latter. isaiahbjork/rebot-teleop's independent MJCF of the same arm agrees with the derived jaw frames; nothing here depends on it.

Not verified, and not verifiable without hardware: everything a headset touches, unchanged from #900 — plus the reBot ghost has never been seen through one. Its placement is derived rather than measured, so what is untested is whether a 190 mm gripper reaching 250 mm out of the fist is usable, not whether it is where the convention says.

Checklist

• I have read and understood the contribution guidelines
• I have run the linter and formatter with SKIP=check-copyright-year pre-commit run --all-files
• I have made corresponding changes to the documentation
• I have added tests that prove my fix/feature works (or explained why not)
• I have signed off all my commits (git commit -s) per the DCO

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 49ed0920-3e44-4b91-9d18-a06bc037f5a7

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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