Which SLAM algorithm for best SLAM

[dinossht]

I've tested rtabmap with livox mid360 lidar and getting good results. But I was wondering if people have experienced better SLAM with other libraries such as Fastlio, LOAM etc. Are there any benchmarks that compares these kind of methods to Rtabmap?

[matlabbe]

If I had time, I would love to try them and compare. As RTAB-Map is kind of a mix of vision and lidar approaches, we may not find a direct comparison in the literature. RTAB-Map is basically using libpointmatcher under the hood to do the heavy ICP work, I would expect RTAB-Map behaving similarly to any other ICP-based approaches. For LOAM-based approaches, they don't do ICP to get transforms, but use geometric features. For a while, it looked like LOAM-based approaches were more accurate/fast, though recently KISS-ICP shown that ICP-based approaches can still be as good.

I updated this page not so long ago: https://github.com/introlab/rtabmap/wiki/List-of-Open-Source-SLAM-projects, some other popular LiDAR approaches are LIO-SAM and FastLio2.

Regards,
Mathieu