matlabbe
matlabbe
Do you tell me that all this time that [example](http://wiki.ros.org/message_filters#Example_.28C.2B-.2B-.29-1) was wrong?! We used to do "sync_qsize=10, sub_qsize=1" by default (like the example) till it was not working anymore on...
We could definitely add a second queue size parameter, maybe called `sync_queue_size` for the synchronizer and keep `queue_size` for the individual subscribers.
With the commit above, we can now set a different queue size for input topics (called `topic_queue_size`) and another one for message_filters synchronizer (called `sync_queue_size`). To keep backward compatibility, `queue_size`...
I [added](https://github.com/introlab/rtabmap_ros/pull/1206) a thread for the odometry nodes, which seems to have fixed the issue there. I tried to do something similar for rtabmap node (a different implementation than yours...
If you reprocess the database (`rtabmap-reprocess`), there should not have any difference between the computers, unless rtabmap is not built with same dependencies. With a rosbag, there could be difference...
Which TUM sequence are you using? and what command did you use?
The variability may be caused by RANSAC on image registration. Here three runs:    ``` rtabmap-report . Directory: ....
If you turn off loop closure detection in localization mode, the robot won't localize in the map. To disable it in mapping mode, an efficient way to disable it is...
From the doc: > Param: GridGlobal/AltitudeDelta = "0" [Assemble only nodes that have the same altitude of +-delta meters of the current pose (0=disabled). This is used to generate 2D...
That is a difficult one, the current code doesn't really support an altitude range for loop closures. The hack approach using GPS altitude would not work as altitude [is actually...