[Bump] Require isaacsim-asset-isolated 6.1 so package:// meshes survive fixed-joint merging

[hujc7]

Description

Moves the standalone URDF/MJCF importers from isaacsim-asset-isolated 6.0 to 6.1, the first line
where merging fixed joints keeps package:// meshes resolvable.

Merging fixed joints relocates the URDF into a scratch directory (/tmp/urdf_import_<robot>_*/),
and the 6.0 importer resolved package://<pkg>/... against that copy. No meshes live there, so
every visual and collision mesh resolved to a missing file while conversion still reported success:
the asset carried its full joint and rigid-body hierarchy and no geometry at all.

The fix is importer-side, in omni_isaac_sim!9476
(nvbug 6606673). package:// now resolves through the configured ros_package_paths mapping, and
falls back to walking up from the source URDF directory when no mapping is given. Nothing on the
Isaac Lab side has to pass a mapping.

Warning

This PR cannot go green yet. See "Blocked on" below — 6.1 is not published, and the 6.1 wheels
built so far are missing modules they import. It is filed now so the pin move is ready the moment
both clear.

Stacked on

Chained on #7192 (jichuanh/usd-exchange-3-rc3). The series is cross-fork, so GitHub cannot
retarget this PR's base to the parent branch; the diff below is the SHA-pinned equivalent.

PR	Own changes
#7192 — usd-exchange 3.0.0rc3 + thread limits	base
this PR	03abad3f3b6...f259b941ae6

Type of change

• Bug fix (non-breaking change which fixes an issue)

Changes

1. pyproject.toml — isaacsim-asset-isolated>=6.0,<6.1 → >=6.1,<6.2.
2. test_source_package_metadata.py — the literal pin assertion mirrors it.
3. source/isaaclab/changelog.d/isaacsim-asset-isolated-6-1.minor.rst — new fragment.

uv.lock is deliberately not regenerated. The pin has nothing to resolve against yet:

$ uv lock
  × No solution found when resolving dependencies ...
  ╰─▶ Because only isaacsim-asset-isolated==6.0.1.0 is available and your project
      depends on isaacsim-asset-isolated>=6.1,<6.2, we can conclude that your
      project's requirements are unsatisfiable.

No other pin moves. Every [tool.uv].override-dependencies entry stays: the isaacsim extra
remains 6.0.1.0, so the overrides that exist to beat its transitive pins (mujoco~=3.11.0,
newton-usd-schemas>=0.4.1, typing-extensions, websockets, coverage) are still load-bearing.
usd-exchange==3.0.0rc3 also stays put — every consumer in the 6.1 chain requires only
usd-exchange>=2.2.2:

Package	usd-exchange requirement
isaacsim-asset-isolated 6.1.0a60	>=2.2.2
urdf-usd-converter 0.3.2	>=2.2.2
mujoco-usd-converter 0.5.0	>=2.2.2

The importer config surfaces are unchanged — urdf/impl/config.py and mjcf/impl/config.py are
byte-identical between 6.0.1.0 and 6.1.0a60 — so UrdfConverterCfg / MjcfConverterCfg need no
adaptation.

Validation

Measured kit-less on anymal_d_simple_description (18 meshes) against an Isaac Lab checkout that
forwards cfg.ros_package_paths unchanged, i.e. with no Isaac Lab-side workaround, so the
importer alone is under test.

#	merge_fixed_joints	ros_package_paths	6.0.1.0	6.1.0a60
1	True	unset	0 B / 0 meshes	375,835 B / 38 meshes
2	True	mapping given	376,109 B / 38	375,835 B / 38
3	False	unset	376,109 B / 38	375,835 B / 38
4	False	mapping given	376,111 B / 38	375,835 B / 38

Row 1 is the defect; the other three were already working and stay working.

The nvbug's own QA step, with no mapping supplied:

python scripts/tools/convert_urdf.py \
    anymal_d_simple_description/urdf/anymal.urdf out/anymal.usd \
    --merge_joints --viz newton

Newton Viewer reports Bodies 13, Shapes 55 (visual 38 / collision 17), Joints 13 and the robot
renders fully textured — matching the figures the bug cites for the working case. Zero
could not be parsed / No such file or directory mesh warnings remain.

The fix was also confirmed directly against the published wheels: loading urdf_utils.py out of the
a59 and a60 artifacts and running the relocation shows package:// unresolved on a59 and resolved
to an on-disk path on a60, with and without an explicit mapping.

Blocked on

1. 6.1 is not published. Public PyPI and pypi.nvidia.com carry only 6.0.1.0 for both
   isaacsim-asset-isolated and isaacsim-robot-schema. 6.1 GA is scheduled for 08/25.

2. The 6.1 wheels built so far cannot import. isaacsim-asset-isolated 6.1 stopped shipping
   modules its own code imports — from a55 onward on develop and rc7 onward on release/6.1,
   including rc9, the current GA candidate. Last good builds are a54 and rc6.

   Missing from the wheel	Imported by
   isaacsim/asset/importer/utils/ (10 modules)	urdf/impl/converter.py:28, mjcf/impl/converter.py:27
   isaacsim/asset/transformer/{__init__,extension,manager,models,rule_interface,utils}.py	transformer/rules/extension.py:37

   A clean install therefore fails at import isaacsim.asset.importer.urdf with
   ModuleNotFoundError: No module named 'isaacsim.asset.importer.utils'. Both extensions still
   exist in the Isaac Sim source tree, and no other published wheel provides them in importable
   form, so this is a packaging gap rather than a refactor Isaac Lab can adapt to. The Kit path is
   unaffected — Kit serves isaacsim.asset from its extension roots.

   Separately, isaacsim-robot-schema 6.1.0a60 ships dist-info only (6.3 KB vs 54 KB at 6.0.1.0,
   no usd/schema/isaac/), which makes the Make Robot Schema transformer rule fail during
   conversion so the converted asset gets no IsaacRobotAPI.

   The validation above was obtained by restoring the missing modules from a54 into the
   environment; it exercises the real 6.1 URDF importer, but is not a pure clean-wheel install.

Once fixed 6.1 wheels are published, this needs only uv lock and a re-run of the table.

Checklist

• I have read and understood the contribution guidelines
• I have run the pre-commit checks with ./isaaclab.sh --format
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• I have added a changelog fragment under source/<pkg>/changelog.d/ for every touched package
• I have added my name to the CONTRIBUTORS.md or my name already exists there

[hujc7]

fixed in upstream