Can I load a map and localize the camera and pose within the map? about rtabmap

[matlabbe]

Breaking down your message:

• "possible to load a map, then localize the camera and its pose within that map": yes
• "update the map accordingly": yes, you would have to manually switch between localization mode and mapping mode after the camera is localized, or start in mapping mode with parameter Rtabmap/StartNewMapOnLoopClosure to true (this will do the "Append Mode").
• "Furthermore, is it possible to extract and store the user position path (keyframes)? My ultimate goal is to store the keyframes as nodes (storing coordinates) with edges between them": Do you mean saving a "user graph" during localization mode? No, you would need to save the poses on the side yourself with the current poses computed by rtabmap.
• "is april tag available for localization?": yes, either by using external node like apriltag_ros or by enabling RGBD/MarkerDetection (if rtabmap is built with opencv's aruco module) and setting Marker/CornerRefinementMethod=3 and Marker/Dictionary=20 for 36h11 apriltag family.

cheers,
Mathieu

[matlabbe]

Do you use ROS? It will be easier to do what you want with the ROS package.

1. See #1178 for the first question. The optimized poses are saved when rtabmap is closed:
   

   rtabmap/corelib/src/Rtabmap.cpp

   Line 518 in f56875d

   _memory->saveOptimizedPoses(_optimizedPoses, _lastLocalizationPose);

2. See localization mode in ROS. In c++, if you look at the RGBD example, add your database path in Rtabmap's constructor and add parameter Mem/IncrementalMemory=false to the params map.
   

   rtabmap/examples/RGBDMapping/main.cpp

   Lines 200 to 205 in f56875d

   	ParametersMap params;
   	//param.insert(ParametersPair(Parameters::kRGBDCreateOccupancyGrid(), "true")); // uncomment to create local occupancy grids
   	// Create RTAB-Map to process OdometryEvent
   	Rtabmap * rtabmap = new Rtabmap();
   	rtabmap->init(params);

[kenhuang1964]

Thank you so much @matlabbe!

[kenhuang1964]

Hey @matlabbe, thank you for your help earlier. I have a few more questions if you don't mind:

1. Do you have any recommendations for how to extract the poses computed by rtabmap without having to use the GUI? Also, when are optimized poses saved?
2. How can I load a map and localize the device within that map?

Thanks in advance!