Relocalization in mapping mode in lidar only system

[cdiaz-ruiz]

I have some questions on re-localization using a lidar only based system (lidar, imu, wheel encoders) -- no cameras. I have successfully been able to begin a mapping session and create a .db, then launch the system in localization mode and recover localization from the database ( I re-launched from where I had stopped the system previously). But when I launch the system in mapping mode I dont seem to see the pose be relative to the previous map. If I am trying to continue running on mapping mode from a previous database and have the system re-localize to the prior session (presuming I keep the robot near where it was last turned off), are there particular parameters to be set to get this to work?
I dont use -d (to ensure using database)
In localization_mode I set
"Mem/IncrementalMemory": "false"
"Mem/InitWMWithAllNodes": "true"
In mapping mode I set:
"Mem/IncrementalMemory": "true"
"Mem/InitWMWithAllNodes": "false"

[cdiaz-ruiz]

I understand that global loop closures are not supported without cameras. But am wondering if I can at least recover in mapping mode if I begin from where we last left the robot upon shutdown. I am using my own odometry and using tf way for rtabmap to retrieve pose.

[matlabbe]

That should be supported since that pull request introlab/rtabmap#1644. The ros binaries have not been updated yet with it, so you would need to rebuild rtabmap and rtabmap_ros from source.