LiDAR SLAM without overlap between consecutive scans

[pablosecco]

Hello,

I was planning to apply SLAM over a MMS-point cloud but I realised that there is no overlap between consecutive scans (from consecutive poses of the sensor).

This is due to the fact that the LiDAR sensor used rotates around a horizontal axis, resulting in vertical scans (vertical lines parallel to each other) without overlap between them. However, the distance between the lines can be very short, e.g. around 1 cm when the MMS moves slowly.

In addition to this, every survey consist of 4 runs which are processed together and averaged. So, every area is revisited 3 times. In this sense, there is certainly overlap, but not among consecutive scans.

I was wondering if it could still be a way to perform SLAM in this situation.

Regards,
Pablo

[matlabbe]

I am not aware how scans are assembled in this technique. How the distance and height between two scans are computed if the don't overlap? Is it based on vehicle velocity?

In RTAB-Map, is is possible to feed an external odometry to assemble scans. If the movement estimation is relatively accurate over a couple of meters, some scans could be assembled and fed to RTAB-Map. On loop closures, instead of having only one ray to register, there could multiple rays instead. However, motion error inside assembled scans would not be corrected.

cheers,
Mathieu