mapping info

[zlacelle]

Sorry if this info exists somewhere, but I've been unable to find it.  If it does, please point me to it and I'll start reading!

As far as I can tell, there are a few map topics published:
* cloud_map - This includes PointCloud2 of the full map, as well as separated into ground and obstacles.
* octomap - This does normal OctoMap mapping (nicer ray-tracing)
* proj_map - This publishes a 2D flattened map (DEPRECATED for grid_map)
* grid_map - This publishes a 2D flattened map
* scan_map - This publishes a pointcloud of obstacles (DEPRECATED for cloud_map)

My questions:
1) proj_map vs grid_map - Do they use different algorithms, or simply different topic names?  They seem identical.
2) proj_map flattens obstacles to the ground. Does it use Z=0 ground plane, or does it use the ground pixels calculated from the Z normals?
3) scan_map vs cloud_map - Do they use different algorithms, or simply different topic names? They seem identical.
4) Is the pipeline disparity/RGBD -> cloud_map -> grid_map / octomap?

Thank you!

[matlabbe]

Hi,

1) grid_map and proj_map are now exactly the same topic, old name "proj_map" kept for backward compatibility.
2) grid_map is computed depending on "Grid" parameters:

$ rtabmap --params | grep Grid/
...
Param: Grid/NormalsSegmentation = "true"                   [Segment ground from obstacles using point normals, otherwise a fast passthrough is used.]

If "Grid/NormalsSegmentation" is true (default), normals are used to segment the ground. If false, passthrough filter is done on z plane, with maximum height of "Grid/MaxGroundHeight" (i.e., everything under "Grid/MaxGroundHeight" is ground).

3) Same as 1), they are same topic.
4) Not exactly. For each node added to map's graph, a local occupancy grid is created using "Grid/" parameters. For grid_map, these local occupancy grid maps (2D or 3D) are assembled depending on the current graph, creating the global occupancy grid. If Grid/3D is true, obstacle and ground points are projected on z plane before assembling grid_map. For cloud_map, local occupancy grids are assembled directly (one point per cell). For OctoMap, each 3D local occupancy grids ("Grid/3D" must true) are appended to OctoMap, doing ray tracing for each of them. So it more like:
Disparity/RGBD -> local occupancy grids (2D or 3D) -> grid_map / cloud_map / octomap

cheers,
Mathieu