Ego localization mode for depth sensor (and IMU) in known environment

[trackexp]

Hello rtabmap community,

I'm currently investigating whether rtabmap is suitable for what I'm trying to do.

My goal is to use a ToF depth sensor (raw point cloud without any color information) for localization in a known environment to achieve inside-out tracking. The environment is known from either a previously generated map or from a 3D CAD model, assumed to be static and within a range of 3 meters. There won't be much variation in position, but a lot of variation in the orientation of the device.
The plan is to integrate the depth sensor in an already existing sensor fusion solution and fuse it with IMU data. With the depth sensor data, the drift of the IMU only 6dof tracking can be compensated. The depth sensor can be easily integrated in ROS.

As far as I understand, rtabmap can be used in a localization mode which uses the loop closure detection mode. I'm not sure, though, whether this is the right method to achieve 6dof localization with the data frame rate of the depth sensor (e.g. 5fps, 15fps, 30fps).

It would be very helpful, if you can give an answer, whether rtabmap is suitable for localization with raw point cloud data as stated above.

Thank you in advance!

Best regards,
Tobias

[matlabbe]

Hi,

Your depth sensor should have a large field of view. Then similar to this post, it could be possible to create a map of the workspace, then switch to localization mode (assuming that the camera is starting always at the same initial position in the map).

cheers,
Mathieu