search

hku-mars / Point-LIO

Other
794 stars 132 forks source link

Fast-Lio2 Performs Better #102

Open celikemir opened 1 month ago

celikemir commented 1 month ago

Hi,

I have several bag files recorded at different altitudes with a UAV, such as 50m, 100m, etc. The LiDAR is positioned at nadir, pointing downward. There is significant odometry degradation, particularly in some areas. I also tested the same bag files with FAST-LIO2, and its performance was better and closer to the ground truth. My question is: do you have any suggestions for adjusting parameters for outdoor use and fixing these degradation points? (I've already adjusted filter_size_surf, etc.) The odometry completely diverges in the opposite direction and fails to recover.

sharknow commented 1 month ago

Hi,

I have several bag files recorded at different altitudes with a UAV, such as 50m, 100m, etc. The LiDAR is positioned at nadir, pointing downward. There is significant odometry degradation, particularly in some areas. I also tested the same bag files with FAST-LIO2, and its performance was better and closer to the ground truth. My question is: do you have any suggestions for adjusting parameters for outdoor use and fixing these degradation points? (I've already adjusted filter_size_surf, etc.) The odometry completely diverges in the opposite direction and fails to recover.

Can you provide a picture with the results? Thank you!

celikemir commented 1 month ago

Here are the results below, where you can easily distinguish which one diverges from the ground truth. Ground truth, Point Lio, Fast-LIO respectively 50m 50m_pointlio 50m_newparam