search

CathIAS / TLIO

Tight Learned Inertial Odometry
Other
277 stars 69 forks source link

TLIO initialization #21

Closed tflucas closed 3 years ago

tflucas commented 3 years ago

Thank you for sharing the code. I have prepared a dataset and obtained a trained model. When I test the TLIO system by using a new sequence with no ground-truth file, how can I obtain an initial state(R v p ba bg) to start the EKF filter. Besides, how can I rotate raw IMU data in IMU local frame to obtain IMU samples in a gravity-aligned frame?

CathIAS commented 3 years ago

If you do not have ground truth or other odometry system running, make an initial guess as accurate as possible and hopefully the filter will converge. In TLIO, this rotation comes from the filter estimates.

patrickctrf commented 3 years ago

Thank you for sharing the code. I have prepared a dataset and obtained a trained model. When I test the TLIO system by using a new sequence with no ground-truth file, how can I obtain an initial state(R v p ba bg) to start the EKF filter. Besides, how can I rotate raw IMU data in IMU local frame to obtain IMU samples in a gravity-aligned frame?

@tflucas Could you share the files you used to train the model? Lots of people trying to understand the TXT files. Thanks!

CathIAS commented 3 years ago

@patrickctrf @tflucas For txt file formats please refer to the documentation. For further questions please open a new issue.