3D mapping with RGB-DSLAM

[Sophye]

Hi,

My project consist to build a 3D map with RGB-D SLAM, i worked with microsoft kinect and turtlebot2 (hydro distro) i want to know if i can do this task with RTAB-Map and also to understand the difference between RTAB-Map package and RGB-D SLAM v2 package and what launch does make the 3D mapping!

Any help please

[matlabbe]

Hi Sophye,

I'll try to answer this question, but I'm not very familiar with the RGB-D SLAM v2 package. The GUI and the results are somewhat similar to RTAB-Map. So with both, you can get a 3D point cloud of the environment with only a Kinect.

For the integration with a robot, I think RTAB-Map may be easier by its flexibility on which sensors you are using. This page shows some configurations that you can use to create your own launch file for Turtlebot2. For example for Turtlebot, you may want a configuration like "Kinect + Odometry + Fake 2D laser from Kinect". This configuration was used in this video https://www.youtube.com/watch?v=_qiLAWp7AqQ, if it is something like this that you want. A 2D map is also created for navigation.

What is the goal of creating a 3D map in your project? Will it be used for robot navigation? or robot teleoperation? or just for visualization? Well, if it is just for visualization, you don't need a robot at all, just use a hand-held Kinect and compare the results between RGB-D SLAM v2 and RTAB-Map:

(Don't launch them at the same time!)

RGB-D SLAM v2:
$ roslaunch rgbdslam openni+rgbdslam.launch

RTAB-Map:
$ roslaunch freenect_launch freenect.launch depth_registration:=true
$ roslaunch rtabmap rgbd_mapping.launch

For navigation in a 3D map, you will need an octomap. Octomap seems more integrated in RGB-D SLAM v2 than in RTAB-Map. However, RTAB-Map generates a 3D point cloud, this point cloud can be converted to an octomap like in this post.

For teleoperation, the Demos section on rtabmap_ros page shows some examples.

Cheers,
Mathieu