FAST_LIO

FAST_LIO copied to clipboard

Thank you for your open source.I have a problem when running your code，large scale scene(for 4-5km in city no repetition) attitude drift lead to IMU acceleration bias drift(2-5m/s^2).I found that roll and pitch drift at 15deg~20deg,according to the formula (3) in FAST-LIO,which R change will lead to ba drift.I want to know if you encounter this problem？and is there any way to solve it？Any reply is helpful.

Actually FAST-LIO is a LIO not a slam system. for long term accuracy, loop closure or pose graph back-end will help, such as FAST-LIO_SLAM.

Moreover, there are still some place need improves (especially the long-term consistency you mentioned) in current version, we will do them in next version.

Thank you for your reply!If any LIO system for example LIO-SAM will also have this problem？I think that gravity is a global measure for roll and pitch,why that cannot constraint the attitude.

Actually you proposed a very good question. For a consistant system, if the pitch roll is observable, there should be less drift in pitch roll or any observable states. But designing a consistant system is a interesting research topic we are working on.

For liosam or any slam system with loop closure, things will be different. the drift should be low down when loop closed.

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.