Map points created from lidar looks weird..

[dinossht]

When setting the voxel_size parameter to 0.2 meters, I observe that the point cloud exported from the database (created using LiDAR data) appears gridded, with points clustered together at regular intervals. I had expected the point cloud to either contain points spaced approximately 20 cm apart, or to be more uniformly distributed across the map.
Is it expected that the point cloud looks like clusters of points distributed around every 20 cm, rather than a more even spread? Could you explain why this pattern appears?

[dinossht]

More images of the map for context:

[matlabbe]

Could be related by how the voxel filter is creating the grid, by rounding the cells position, that could give many points appearing in same cell when the lidar didn't move more than 10 cm, then jump to cluster 20 cm and so on.

It also depends in which frame the voxel filter is applied, in odom or base_link frame? In base_link frame, I would expect that assembling the resulting points clouds in map frame would spread more the points.

Are those point clouds voxelized by point_cloud_assembler?

[dinossht]

Yes, seems like point_cloud_assembler is the reason for this point distribution. Using the raw lidar data directly to rtabmap_slam node removed this problem.