goldbattle
commented
1 year ago This might be due to your use of the quaternion. We use a JPL quaternion which is the rotation from the GLOBAL to the IMU frame. I recommend you try writing the quaternion x,y,z,w to file, and then processing it. When a JPL quaternion is written in this order, and read as a Halmiton quaternion, it flips the rotation direction to IMU to GLOBAL.
On Sun, Jan 22, 2023 at 11:58 PM Henry Che @.***> wrote:
Hi,
I'm trying to run OpenVINS with EuRoC, and it seems that although the xyz and pitch match well, both roll and yaw differ have a constant offset of about 50 degree compared to ground truth. [image: image] https://user-images.githubusercontent.com/90811347/213965631-4cd9fce2-30a2-4211-93fb-f336310d1572.png [image: image] https://user-images.githubusercontent.com/90811347/213960155-2327bee1-b005-4402-8547-82db3455d252.png
The config file that I used is the exact ones https://github.com/rpng/open_vins/tree/master/config/euroc_mav provided by OpenVINS. My code is here https://github.com/hungdche/open_vins/blob/master/ov_msckf/src/run_illixr_msckf.cpp just in case.
Any help for pointer is greatly appreciated. Thanks!
— Reply to this email directly, view it on GitHub https://github.com/rpng/open_vins/issues/303, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAQ6TYSCVRORCNFX3FLCBLLWTYFXDANCNFSM6AAAAAAUDNJ3NY . You are receiving this because you are subscribed to this thread.Message ID: @.***>
hungdche
Hi,
I'm trying to run OpenVINS with EuRoC, and it seems that although the xyz and pitch match well, both roll and yaw differ have a constant offset of about 50 degree compared to ground truth.
The config file that I used is the exact ones provided by OpenVINS. My code is here just in case.
Any help for pointer is greatly appreciated. Thanks!