Taking almost 3 seconds to update the 2D OGM in a small scene.

[mi]

Hello! :)

So, I've been using your project in a very large scene, and the 2D OGM was successfully updated inside the range of 1 second (grid resolution of 0.05m). Now I'm running in a very small scene, and the update is taking 1, 2, and even 3 seconds.
Since I'm moving the camera around the scene (for example, I show the same point of view about 2-3 times), I'm thinking the delay is because of the (dense) point cloud optimization given a loop detection. Or even the loop detection is causing the delay, but since this is done by another thread and should not take that long when using features for place recognition, I believe this is about map optimization.
So, would be wonderful to hear from you about what do you think is causing this.

Thanks in advance. Att,

[matlabbe]

Hi,

One way to check which part is taking the most time is by looking at the statistics saved in the database. They can be shown in rtabmap-databaseViewer under Statistics panel.

From command line, it is also possible to do so. To show available statistics in the database:

rtabmap-report --stats rtabmap.db

Then to show main steps:

rtabmap-report TimingMem/Add_new_words/ms TimingMem/Compressing_data/ms Timing/Map_optimization/ms Timing/Posterior_computation/ms Timing/Add_loop_closure_link/ms  rtabmap.db

cheers,
Mathieu