Fix Euler rotation order to use ZYX (intrinsic) instead of default XYZ (intrinsic)

[petertheprocess]

Summary

• Fix Euler rotation order to use ZYX (intrinsic) instead of default XYZ to correctly match URDF/MJCF euler angle conventions
• Update rotation settings in MJCFAdapter.js, USDAdapter.js, and InertialVisualization.js

Problem

Three.js Object3D.rotation.set() uses XYZ euler order by default, which differs from the euler order used in URDF formats, and the self-built quaternionToEuler function https://github.com/petertheprocess/robot_viewer/blob/53d3103c2a818d20358fc2bec423303147212fa0/src/adapters/MJCFAdapter.js#L999-L1024.

This caused incorrect orientation of visual/collision meshes and inertial visualizations.

Solution

Explicitly specify 'ZYX' euler order when setting rotations to ensure consistency with robot model format conventions.

Result

Using agilex_piper mjcf from mujoco_menagerie as demonstration

Before (Default XYZ)

After (ZYX)

TODO for improvement

• parse eulerseq tag for mujoco mjcf, for ref: https://mujoco.readthedocs.io/en/stable/XMLreference.html#compiler-eulerseq
• Refactor internal rotation representation in UnifiedRobotModel.js: Switch from Euler angles (origin.rpy) to Quaternions during the model load stage. Storing rotations as Quaternions avoids ambiguity in the rendering pipeline and eliminates the risk of missing the explicit 'ZYX' order in Three.js rotation.set() calls.

More ref for euler order

• Extrinsic & intrinsic rotation: Do I multiply from right or left?
• Euler angles wikipedia

[fan-ziqi]

Thank you for fixing it! Will you continue implementing the changes in the TODO list?

[petertheprocess]

Thank you for fixing it! Will you continue implementing the changes in the TODO list?

For now, I think it’s better to leave it as is. The TODO list requires a bit more refactoring, and I’d prefer to handle that in a separate PR when I have more time.