Quaternion rotation in tf2_sensors_msgs.h about imu_pipeline HOT 6 CLOSED

No, I borked it, my linear algebra is lacking.

ros/geometry2#78

Needs to be fixed here, PRed up to tf2, and a separate node needs to exist specifically for doing the ned->enu conversion as (poorly) implemented here.

@tonybaltovski FYI

from imu_pipeline.

@paulbovbel Hello, I would like to use this package that seems useful to me, as I am just starting with Ros.

I do not understand if there is actually a problem in the way you do the rotations with quaternion, as @kogut pointed out , and also in your pull request to tf2.
Then, should I remove, in line 112 of tf2_sensor_msgs.h, the first quaternion multiplication between the transform rotation and the IMU rotation; that is, i should do only: imu_rot_quat * transf_quat.inverse()?
About the covariance matrix rotation, I looked that it should be rotated as RCR^t (^t: the transpose) where R is a rotation matrix? Could be that you rotated only once?
I do not understand why you used the quaternion instead of rotation matrices in the first place (as you did in the beginning with angular velocity and acc.).

Thank you in advance!

from imu_pipeline.

@ClaudioCimarelli, to be perfectly honest, I remember nothing about how or why anymore. I haven't touched an IMU in about 3 years.

PRs are most certainly welcome if you have the time. A gtest or two to validate the math would probably go a long way!

from imu_pipeline.

Just lost a day on this one! I'll consider sending a PR with the fix and tests if there's chance it'll be merged in a reasonable time frame.

from imu_pipeline.

Done: #15 .

from imu_pipeline.

Fixed by #15. (after almost 7 years :-D )

from imu_pipeline.