Revision history [back]

For it to work most effectively, AMCL requires new information from each lidar scan it ingests. If the robot is not moving, feeding AMCL the same scan over and over will make the algorithm perform much worse.

AMCL is designed to handle Global Localization for difficult cases: imagine multiple identical rooms, each with a doorway to a shared hallway. If the robot is inside one of the rooms, there is no way to determine which room it's in from just the lidar scans - the robot will have to move into the hallway to determine a single solution. This is why AMCL wants "new" scans.