Fix trimesh object tilt limit when max_yx_angle is in radians

[Patrick-SCH03]

Description

make_box, make_cylinder and make_cone in isaaclab/terrains/trimesh/utils.py document max_yx_angle as "the maximum angle along the y and x axis", with degrees choosing the unit. They draw a random rotation, take its "zyx" Euler angles (the x angle lies in [-pi, pi]), and scale the y/x angles:

if degrees:
    max_yx_angle = max_yx_angle / 180.0
euler_zyx[1:] *= max_yx_angle

With degrees=True this caps the x angle at max_yx_angle. With degrees=False the angles are multiplied by the raw radian value, so the cap is exceeded by a factor of pi. Sampling 5000 objects with a 30 degree cap:

call	max |x angle|	max tilt
max_yx_angle=30.0, degrees=True	29.99 deg	32.4 deg
max_yx_angle=np.deg2rad(30), degrees=False	94.25 deg	94.2 deg

so objects "capped" at 30 degrees can tip past horizontal. Box, cylinder and cone behave the same.

This divides by pi in the radians branch, so both units give the same cap. The degrees=True path is unchanged. No in-repo caller passes degrees=False.

(Separately, the y angle only reaches half the cap on both paths because its Euler range is [-pi/2, pi/2]; I left that alone since changing it would alter existing terrains.)

Type of change

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

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 (do not edit CHANGELOG.rst or bump extension.toml — CI handles that)
• I have added my name to the CONTRIBUTORS.md or my name already exists there

Testing

• Added test_object_max_yx_angle_units to test/terrains/test_terrain_generator.py, parametrized over the three functions: with the same seed, a 30 degree cap in degrees and in radians must give the same mesh. It fails for all three on develop and passes with the fix.
• The whole test_terrain_generator.py passes locally (CPU, with Isaac Sim imports stubbed out).
• Ruff v0.14.10 (the pre-commit pin) check and format are clean on the changed files. I ran it directly because the full ./isaaclab.sh --format environment isn't available on my Windows machine.

Release backport

• Backport this pull request to the active release branch after it merges into develop

[greptile-apps]

Confidence Score: 5/5

[Medium risk] Fixes terrain object tilt angle unit conversion bug.

The PR appears safe to merge.

Summary

The PR normalizes radian-valued tilt limits for trimesh boxes, cylinders, and cones to match the existing degrees behavior.

• Adds a seeded regression test comparing meshes generated with equivalent degree and radian limits.
• Adds a changelog fragment describing the fix.

_{Reviews (1) · Last reviewed commit: "Fix trimesh object tilt limit when max_y..."}

[isaaclab-review-bot]

Isaac Lab Review Bot

The patch consistently normalizes radian max_yx_angle values by π in all three trimesh object helpers, matching the existing degree-path scaling. The parametrized regression test and changelog fragment cover the corrected behavior.

• Design and architecture: The change preserves the existing normalize-then-scale Euler-angle design and applies the same targeted correction to make_box, make_cylinder, and make_cone without adding dependencies or abstractions.
• API: Public signatures and defaults remain unchanged. The patch corrects the degrees=False behavior to match the documented unit contract, while preserving the established degrees=True path and recording the user-visible fix in the package changelog.
• Implementation: The scaling is consistent: 30 / 180 and deg2rad(30) / π both produce the same multiplier. The test reseeds before each call and compares observable mesh vertices across all three helpers, including consistent random section counts for cylinders and cones. Reliance on SciPy's default global random source is a minor test portability consideration, not a demonstrated defect.

No blocking issues. No inline issue met the actionable-evidence threshold; the assessment above records the review feedback.

Automated review; human maintainers own approval decisions.

[StafaH]

run-ci

[isaaclab-bot]

Backported to release/3.0.0 as 82fe9b5.