[hougantc-nvda]

Description

Review stack · 3/5: #8787 → #8788 → #8789 → #8790 → #8791

Review only this PR’s changes. The PRs target main; the comparison isolates this layer from its predecessors.

Enable the registered GR1T2, G1 locomanipulation and fixed-base G1 tasks to run with kitless Newton or CPU OvPhysX. Explicit physics selectors choose matching task presets and USD variants; omitting the selector preserves Kit defaults.

• Task ownership: robot mappings, action conditioning, model assets and backend tuning stay in isaaclab_tasks. Teleop consumes those configurations through the existing adapter.
• Joint ordering: preserve the configured Pink joint order so hand targets reach the intended joints.
• Shared assets: extract the steering-wheel helper used by GR1T2 and G1, retaining graspable rim/spokes and valid collision shapes without a cross-task import.

The existing G1 head camera remains available at this PR boundary; wrist cameras and presentation changes belong to #8790.

Task selector tests remain independent of teleop imports. Teleop preparation and the native integration harness stay in isaaclab_teleop.

See registered tasks and physics selectors.

Dependencies: compatible Isaac Capture OVXR packages and the provider upgrade in #8858. Native-provider defaults are unchanged.

Type of change

• New feature
• Bug fix
• Documentation update

GR1T2 presets use the supported joint_effort_limit field with the task-specific tuned values.

Testing

Results cover the cumulative five-PR stack on main 6ef40d19a6f8d3770f86b5c4df3f8ccfb6a25f95. Kitless checks use Isaac Capture ae48621f1cc13054c2246210d8f92d30e27fe354 built from source using Python 3.12; Kit checks use Lab's upstream locked Capture selection. They are not independent full-repository results for each PR.

• 1,039 cases across 51 selected Lab/Kit files passed, with no remaining failures or skips. Coverage includes ordinary rendering, native cross-GPU publication, graph capture, startup interruption, failed-drain ownership, retargeting, task selectors, and 140 Kit-selected cases including launch/replay and Scene UI.
• Five Omen IWER scenes passed stereo/task-object inspection, meaningful input, Reset and normal mirror-close shutdown: GR1T2 CPU and split-GPU Newton, G1 locomanipulation CPU, fixed-base G1 Newton and SO101 CPU. Humanoid PiP was checked; SO101 uses mirror only. Reconnect/resumed input was checked for G1 and GR1T2 Newton.
• The ordinary Kit headless Franka XR path passed stereo scene, motion, Reset and reconnect checks. It ended by deliberate owner interrupt after client disconnect; its detached CloudXR service was explicitly stopped afterward.
• Formatting, changelog checks, warning-as-error Sphinx and the kitless dependency/selected-extra checks passed. The installed Capture wheel and staged package files were verified identical.

Limits: these are synthetic IWER checks, not physical-headset acceptance or full task completion. Cold renderer attach, intermittent GR1T2 first-Play recovery and SO101 cube retention remain open. Kit's upstream dependency selection reports nine metadata inconsistencies, including NumPy/Numba; passing the selected Kit tests does not qualify that entire dependency graph. Full-repository, hosted-CI and arm64 coverage are not claimed.

See #8791 for the matched video-rate measurements.

Checklist

• I have read and understood the contribution guidelines
• I have run the pre-commit checks with uv run isaaclab --format
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have validated the changed behavior with focused coverage and applied the test-audit criteria to any test changes
• I have added a changelog fragment under source/<pkg>/changelog.d/ for every touched package (do not edit CHANGELOG.rst or bump package versions — CI handles that)
• I have added my name to the CONTRIBUTORS.md or my name already exists there

The warnings checkbox remains open: dependency warnings and the separate Kit metadata audit are unresolved.