Need Clarification on RTAB-Map settings and their effects

[AyushBharadwaj]

Hello @matlabbe 👋,

I would like to get a better understanding of the various RTAB-Map settings and how they affect overall performance and mapping quality. Specifically, I’m looking for guidance on:

1. What each key setting or parameter does.
2. How some parameters influence others (for example, those related to depth filtering, loop closure detection, and voxel grid resolution).
3. Recommended or commonly used configurations for different scenarios such as indoor room scanning, object scanning, or large-area mapping.

It would be very helpful to have a clearer picture of how these settings interact and what the best practices are for adjusting them.

Thank you for your time and support! 🙏

— Ayush

[matlabbe]

Going into details for every possible applications and sensors would be too long to do. The general guide for robotics is from the paper, like this one to start. For loop closure detection, see this paper. Other papers here (e.g., multi-session mapping, memory management, illumination invariant visual localization): https://introlab.github.io/rtabmap/#overview

For 3D scanning, the default parameters chosen in RTAB-Map iOS app are appropriate for most room-size scanning.

The easiest way for me to give more personalized tips is for people sharing a rtabmap.db database they created.