Dose State.qd[3:6] represents the angular velocity in the local body frame?

[hdadong]

Is it correct that in State.qd:

• State.qd[0:3] represents the linear velocity in the world frame, and
• State.qd[3:6] represents the angular velocity in the local body frame?

Are there any document for these?

https://github.com/google/brax/blob/main/brax/generalized/base.py#L71

[btaba]

Edit: q and qd are analogous to qpos and qvel in MuJoCo land. If you're referring to a free joint, qvel[0:3] will be the linear velocity in the global frame and qvel[3:6] will be the angular velocity in the local body frame

[hdadong]

No, you could find in the mujoco documentation, the qvel[3:6] is the angular velocity in the local body frame.

https://mujoco.readthedocs.io/en/stable/overview.html#floating-objects:~:text=However%2C%20the%20rotational%20velocities%20of%20a%20free%20joint%20are%20in%20the%20local%20body%20frame.

some discussion here:
google-deepmind/mujoco#691 (comment)

You can check this:

• The result of inv_rotate(pipeline_state.xd.ang[0], quat) is equal to pipeline_state.qd[3:6].
• Since xd.ang represents the angular velocity in the world frame, the value in pipeline_state.qd[3:6] is expressed in the local frame.

[btaba]

Ok I stand corrected! Nice digging @hdadong

[hdadong]

In the forward function in this code: you transform the qd[:, 3:6] to the xd, so the qd[:, 3:6] is in the local frame.

https://github.com/google/brax/blob/main/brax/kinematics.py#L51
https://github.com/google/brax/blob/main/brax/kinematics.py#L92

[hdadong]

Any news about this?