the same param for indigo branch and master branch, but different effect

[mantouRobot]

Hi~
      I use different branch of rtabmap and rtabmap_ros, but have the same params for localization mode. Let me confused is that the same params but different effect. The params for the indigo branch is ok but not good for the master branch. Any advice?
      Thank u so much~

[matlabbe]

You should always use the same branch for rtabmap and rtabmap_ros. The master branches have the latest updates. Normally, if a parameter has changed, you would be noticed with a warning or error in the terminal. You can post here these warnings/errors if don't know what to do.

cheers

[mantouRobot]

Hi~
     How can i know the map is large enough?
     Thank u~

[matlabbe]

Large enough for localization? ~30 locations for my apartment is okay. I generally do a 360 scan of the environment, find a loop closure to correct the map globally then stop. On localization, I move the camera in the same trajectory until it localizes.

[mantouRobot]

Hi~
      "move the camera in the same trajectory until it localizes". The "same trajectory" u mean is the trajectory that the camera for the mapping period produce? I found that when I not in the same trajectory for it localization, the localization result is not good as in the same trajectory. May be the SURF will be more better for anywhere localization?
      Thank u~

[matlabbe]

Hi,

The camera must be close to a pose in the graph to localize. Otherwise, when you move outside the trajectory mapped, it is odometry until a new localization can be done (when re-crossing the map's trajectory). If you want to localize more often in different camera positions, it is better to map the area with the maximum number of positions/orientations to better "learn" the area. SIFT/SURF features may produce better localization hypotheses.

cheers