AMCL

A probabilistic localization system that uses particle filters to track a robot's 2D pose against a known occupancy grid.

AMCL (Adaptive Monte Carlo Localization) is the standard for 2D robot positioning within the ROS and ROS 2 ecosystems. It utilizes a KLD-sampling particle filter to estimate a robot's pose against a static map by processing LIDAR scans and odometry data. The algorithm maintains a cloud of candidate poses (particles) that converge as the robot moves: delivering centimeter-level accuracy in environments like warehouses or hospitals. It effectively handles the kidnapped robot problem by injecting random particles when localization quality drops. Most modern deployments use the nav2_amcl package to balance CPU overhead with high-precision tracking.

https://navigation.ros.org/configuration/packages/configuring-amcl.html