Streaming RTABMAP into RVIZ

[jacksonkr]

I'm able to view the RTABMAP map data in RVIZ using /rtabmap/mapData but that only brings in point cloud data and as more data is collected the RVIZ display gets really messy because it's not interpreting the transformations of the points correctly. The only way to reset to a clean map is using MapCloud > Download map.

Is there a better way to get a constant stream of the model produced by rtabmap from a remote device? Is it possible to get mapData for the 3d model rather than the point cloud data?

[matlabbe]

Hi,

I am not able to reproduce the bug, which rtabmap version are you using?

I didn't test on remote, but the following would do the same thing:

$ roslaunch openni2_launch openni2.launch depth_registration:=true
$ roslaunch rtabmap_ros rtabmap.launch rtabmapviz:=false rviz:=true

The point clouds are moved when a loop closure happens.

Note that if you reset the map, calling "$ rosservice call /rtabmap/reset", you should uncheck/check the MapCloud plugin in RVIZ to refresh its cache, otherwise it would show old point clouds instead of the new ones.

cheers,
Mathieu

[jacksonkr]

RVIZ 1.11.14

I found that my master was using an old version of rtabmap. I updated rtabmap (from source) as well as rtabmap_ros (from source). Both robot and master are now 0.11.16. As you mentioned, using rviz for /mapData works perfectly fine locally. Doing the process remotely is a different story..

Master:

$ roscore

$ rosrun rviz rviz

Robot:

$ roslaunch freenect_launch freenect.launch depth_registration:=true

$ roslaunch rtabmap_ros rtabmap.launch rtabmapviz:=false rviz:=false rtabmap_args:="--delete_db_on_start"

Screenshots:

Fresh start of rtabmap with rviz:


Rviz display after moving the camera around:


Rviz display after MapCloud > Download map


Might it have something to do with freenect? openni / openni2 won't run for me or I would try those.

[matlabbe]

The point clouds are added one over the other. Make sure your fixed frame in RVIZ global settings is set to /map.

[jacksonkr]

Thanks for your patience! Of course that's what it was..