Submap for PGO? about fast_lio_lc HOT 8 OPEN

Hi, it is the submap form.

If a loop closure is found, the registration result between their submap will be added to the pose graph.

from fast_lio_lc.

Thanks, great work~

from fast_lio_lc.

Hi, @yanliang-wang

There are still some questions that i want to ask you for help.

1. Did u notice the missing blocks in the scan? In my experiments with VLP16 and your gif as well, i saw some blocks of missing points in some rings of one scan. During the odometry, the positions of the missing block vary. Is this phenomenon caused by motion undistortion(certainly not by occlusion)? But still, i think it is weird in fastLIO.
   
   

2. When i check the raw time field in scan produced by velodyne_driver, it is negative. Do u know the reason for this?

from fast_lio_lc.

Hi, @yanliang-wang

I noticed you construct scancontext just using keframes instead of submap accumulated by kfs?

from fast_lio_lc.

Hi, @yanliang-wang

I noticed you construct scancontext just using keframes instead of submap accumulated by kfs?

FAST_LIO_LC/PGO/src/laserPosegraphOptimization.cpp

Line 678 in d059255

scManager.makeAndSaveScancontextAndKeys(*thisKeyFrameDS);

Yes, this part is from FAST_LIO_SLAM. I didn't change it. But it doesn't matter if you only use the radius-based loop closure.

from fast_lio_lc.

You were saying you also use frames to detect loop but match submaps to close the loop instead?

from fast_lio_lc.

You were saying you also use frames to detect loop but match submaps to close the loop instead?

Yes, is there any problem?

from fast_lio_lc.

Hi, it is the submap form.

If a loop closure is found, the registration result between their submap will be added to the pose graph.

Of course no, I misunderstood. I thought you use submap for loop detection as well. Thanks～

from fast_lio_lc.