amcl with rotated map

edit

I find this an interesting question. I have used rotated maps in the past and I did not experiment a great localization error.

I also rather prefer to have an aligned map because of tracking of pose components is much simpler than taking into account the rotation of the map. Though, not always is possible to create a live map that would satisfy this.

I think first you should evaluate effect of rotation of your map on the localization. It depends on grid scale, map size,...

In map_server documentation ( http://wiki.ros.org/map_server ) it is told that the .pgm file is processed by SDL_image ( http://www.libsdl.org/projects/SDL_im... ). Map server converts pixel color into occupancy value, so I guess that if you modified the map in a photo editor (like InkScape/Gimp/...) as long as size matches with metadata yaml file I do not see any problem. I guess that what is more likely to happen when editing manually is that your origin point would move sligthly to the one you record. I did not try but I will and update with conclusions.

Maybe it is just that I am not aware of inner effects on AMCL, but I do not see problems with working with rotated maps more than is more difficult to perform tests/pose debugging. You can always make use of AMCL parameters to handle initial mean hypothesis condition as launch param or directly on the ParameterServer:

~initial_pose_x (double, default: 0.0 meters)

Initial pose mean (x), used to initialize filter with Gaussian distribution.

~initial_pose_y (double, default: 0.0 meters)

Initial pose mean (y), used to initialize filter with Gaussian distribution.

~initial_pose_a (double, default: 0.0 radians)

Initial pose mean (yaw), used to initialize filter with Gaussian distribution.