About 2D SLAM libraries in MRPT project (like rbpf-slam & icp-slam libraries using 2D LiDAR), what happens when environment changes or people walking around the robot? The library detects these changes in environment and eliminates them? Or applies changes into the map during navigation? Or they are simply considered as sources of errors in position calculation?
What is MRPT approach to environment changes during robot navigation?
52 views Asked by Mazdak At
1
If your question scope is SLAM itself only: yes, the changes are integrated into the map as they occur.
The way this happens depends on the metric map type: for pointclouds, you will end up with many "noisy" areas where the dynamic objects moved. For occupancy grids, if there are more observations detecting a given area as "free" than "occupied", in the end it will be marked as "free" (and the other way around).
Obviously, having many dynamic objects during mapping could degrade the quality of localization during SLAM, hence that of the map itself.