Revision history [back]

AMCL provides the "global_localization" service for this use case (see AMCL wiki page). It disperses particles randomly in the given map and hopefully finds the correct robot pose after the robot moved around a little bit, pruning "bad" particles. You´ll have to test how well that works for your scenario, though. I haven´t played around with it much, but I´d expect that huge and redundant looking environment (e.g. with lot of places looking the same) will be a challenge. If you are able to set the initial_pose, that definitely would be the more reliable method.