Simultaneous Localization and Mapping

A robot building a map of an unfamiliar environment while simultaneously figuring out where it is on that map.

Simultaneous localization and mapping refers to a robot, on entering an unfamiliar environment, building a map of it while also estimating its own pose within that map at the same time. The difficulty is that the two problems depend on each other: localization needs a map, but mapping needs to know where you are. The field traces back to Randall Smith and Peter Cheeseman's 1986 work on spatial uncertainty and to Hugh Durrant-Whyte's group's research in the early 1990s; the acronym SLAM itself first appeared in a survey paper at the 1995 International Symposium on Robotics Research (ISRR). Classic solutions include the extended Kalman filter, particle filters, and graph optimization; approaches are also grouped by sensor, into lidar SLAM, visual SLAM, and so on, with open-source implementations including Google's Cartographer and ORB-SLAM3. It underlies navigation in robot vacuums, warehouse AMRs, and AR headsets.

ExampleA robot vacuum entering a new home for the first time uses its lidar to scan out a floor plan while moving, simultaneously computing its own coordinates on that map in real time; subsequent cleaning runs then plan their routes using this map.

Also called

SLAM

Related

Visual SLAM · LiDAR SLAM · Loop Closure Detection · Visual-Inertial Odometry · Occupancy Grid Map · Adaptive Monte Carlo Localization

Sources

Wikipedia: Simultaneous localization and mapping
Cartographer documentation
Durrant-Whyte & Bailey, Simultaneous Localisation and Mapping (SLAM): Part I (IEEE RAM 2006)，含 SLAM 缩写 1995 年在 ISRR 提出的说明 (Chinese)