Robotics
SLAM
ActivePublished
Key
innovation
Simultaneous, mutually dependent estimation of map and trajectory without a prior map or external positioning.
Category
Robotics
Abstraction level
Paradigm
Operation level
Robot controlSystem
Use cases
Autonomous navigation of mobile robots and AMRsCleaning and warehouse robotsDrones and autonomous vehiclesAR/VR and 3D reconstruction
How it works
Sensor data provides incremental motion measurements (odometry) and observations of environment features. SLAM maintains an estimate of the trajectory and map, updating them with each measurement; detecting a previously visited place (loop closure) corrects accumulated drift. Modern systems use pose-graph or factor-graph optimization.
Problem solved
A robot moving through an unknown, GPS-denied environment (e.g. indoors) needs both a map and its position — yet each depends on the other. SLAM solves this chicken-and-egg problem by estimating both at once.
Key mechanisms
Odometry and scan/feature matching
State estimation (filter or graph optimization)
Loop closure
Map management (e.g. submaps)