Unexpected artifacts when only perform mapping with gt pose. about pin_slam HOT 2 CLOSED

Thanks for your interest in our work.
I didn't notice this issue before and thanks for pointing it out.
After inspecting the data, I found out this issue could be mainly due to the unsmooth "groundtruth" poses of the KITTI dataset.

As you can see in the above figure, the red trajectory is the one estimated by PIN-SLAM and the black trajectory is the "ground truth" GNSS-INS pose provided by KITTI. There's a jump on the Z-axis for the black trajectory. A similar situation can be found at the beginning of KITTI sequence 08. In some other sequences, you can clearly see there is some inconsistency in the trajectory at a revisited region on the Z-axis.

I suspect this unsmooth "groundtruth" poses is the cause of the unexpected artifacts. With the smooth pose estimated by PIN-SLAM, the artifacts would not appear and most of the dynamic objects can be removed by the free space sampling via raycasting.

I hope this help to answer your concern.

from pin_slam.

Hi @YuePanEdward, Thanks a lot for your reasonable replay! I believe it does be the true reason resulting in these artifacts.

from pin_slam.