Mapping with multiple rgbd cameras

[Jonipol]

Hello,

I am working on a project where we use multiple rgbd cameras for mapping and localization. So far I have managed to get the mapping to work with all cameras by building rtabmap and rtabmap_ros from source and setting the parameter DRTABMAP_SYNC_MULTI_RGBD to ON from the CMakeList file. This has significantly increased the build time of our docker image, which was expected as it was mentioned in documentation.

We were wondering if there is another way to do the mapping, using multiple rgbd cameras? That could be possible to do without building from source.

Thank you,
Joni

[matlabbe]

Hi,

RTABMAP_SYNC_MULTI_RGBD is set to OFF by default to avoid problems (too long builds) on ROS buildfarm.

Those files use a lot of RAM to compile, because they are instancing many ROS synchronizers. Not sure how to reduce this. On some computers, reducing the number of parallel builds could help to avoid using the SWAP, thus increasing compilation speed.

Instead of "catkin_make" do "catkin_make -j2" to build maximum of 2 files at the same time. The RGBD sync files require around 2 GB RAM each.

cheers,
Mathieu

[rafachao]

Hi Mattieu,

First of all, thanks a lot for such an amazing software...

We are developing a robot using 2 D435i cameras, one at front, the other one at rear.

Everything goes fine when we are creating the map, but when we switch to navigation, the cameras continue drawing the map, instead of leaving the map "as it is".... When we do this with one camera, and we switch to navigation, the camera doesn't continues drawing the map....

What are we doing wrong?

Thanks for all !!!!!!!!!

Rafa

[rafachao]

Hi Mattieu,

The problem I had using 2 cameras was solved "re-doing" de launch file.

Now I´m testing the performance of rtabmap and ROS using 2 cameras...

Thanks !!!!!