SLAM Loop Closure: How a Robot Fixes Its Own Map
Small errors add up until a robot thinks it is somewhere it is not. Recognising one familiar place lets it correct the whole map.
The robot repeats this loop many times a second. Tap a step to pause on it.
A robot estimates its movement from wheels or cameras, and every estimate is slightly off. Over a long trip the errors pile up, so the map drifts and a loop in the real world looks open on the map.
Loop closure is spotting that the robot has returned to a place it has seen before. That one match gives a strong constraint, and the mapping software adjusts the whole path so the map lines up again.
The map stays consistent over long trips, and the robot can trust where it is.
Drift keeps growing, corridors double up, and the robot ends up lost in its own map.
A wrong match, for example two similar-looking corridors, can warp the map badly. Systems check candidates carefully before accepting one.
What triggers a loop closure?