rotationMatrix() and angles give different results

[libdoron]

rotationMatrix() and angles give different results

[DmitriyKorchemkin]

angle?() methods return rotation angle around corresponding axis that acts as similar as possible (in l2 sense) on the points from corresponding plane. It is not expected to match with Euler angles in all cases except some degenerate ones (rotations purely around one of the axes).

scipy.spatial.transform.Rotation.from_euler supports both intrinsic and extrinsic Euler angles denoted by different case of letters (parameter seq described in https://docs.scipy.org/doc/scipy/reference/generated/scipy.spatial.transform.Rotation.from_euler.html ). If you try scipy.spatial.transform.Rotation.from_euler('XYZ', ...) -- you'll get Rorigin (up to precision of specified values)

[libdoron]

Thanks @DmitriyKorchemkin
Using scipy.spatial.transform.Rotation.from_euler('XYZ', ...) with the angles I get from the 'angle?()' functions, I still don't get the original rotation matrix.
Taking the example from my original post:

yaw_rad, pitch_rad, roll_rad = 0.080531626939773560, -0.184074640274047852, -0.106850937008857727
Rangles = R.from_euler('XYZ', [yaw_rad, pitch_rad, roll_rad], degrees=False).as_matrix()

I now get:

array([[ 0.97749926,  0.10484603, -0.18303689],
       [-0.12094245,  0.98950412, -0.07908558],
       [ 0.17282394,  0.09944303,  0.97991988]])

Which is still not so similar to Rorigin.
I looked over multiple examples and the results I get are still very different.
Still, do you think it is correct to assume that if I need the Euler angles I should stick to the rotationMatrix() function and extract the euler angles from there?

[DmitriyKorchemkin]

I've meant that you need to use scipy.spatial.transform.Rotation.from_euler('XYZ', ...) with the angles you get from Eigen::MatrixBase<...>::eulerAngles(0, 1, 2);.

Seem to match perfectly:

>>> import scipy.spatial.transform as sst
>>> sst.Rotation.from_euler('XYZ', [0.071102283895015717, -0.187578573822975159, -0.100159876048564911]).as_matrix()
array([[ 0.97753477,  0.09823849, -0.18648049],
       [-0.11292146,  0.99114945, -0.06979621],
       [ 0.17797336,  0.08928587,  0.97997628]])

Angles that you get from Sophus::SO3<...>::angle?() will never match Euler angles except some corner-cases, because they're different and somewhat unrelated quantities.

[libdoron]

I checked and you are right - they match. Thanks!