RTAB-Map

An open-source graph-based SLAM library that supports mapping and localization with RGB-D, stereo, or lidar sensors.

RTAB-Map (Real-Time Appearance-Based Mapping) is an open-source SLAM (Simultaneous Localization and Mapping) library developed by Mathieu Labbé and François Michaud at IntRoLab, Université de Sherbrooke, in Canada. It organizes the places a robot has visited into a pose graph, and uses a bag-of-words model — which quantizes image features into “visual words” for comparison — to perform loop closure detection: recognizing “I’ve been here before” adds a constraint to the graph, which is then cleaned up with graph optimization to remove accumulated error. Its memory-management mechanism keeps only a subset of locations active for real-time detection and optimization, so it stays real-time even in large environments over long runs. It supports RGB-D cameras, stereo cameras, 3D lidar, and 2D lidar, and connects to ROS through rtabmap_ros, making it one of the most commonly used off-the-shelf solutions for mobile robot mapping and navigation.

ExampleIn ROS 2, an RGB-D camera runs rtabmap_ros while a robot is pushed around a lab; the result is a 3D point-cloud map and a 2D occupancy grid, handed off to Nav2 for navigation.

Also called

Real-Time Appearance-Based Mapping, rtabmap, rtabmap_ros

Related

Simultaneous Localization and Mapping · Visual SLAM · Loop Closure Detection · Occupancy Grid Map · Common ROS SLAM Packages (GMapping / SLAM Toolbox / RTAB-Map) · Depth Camera

Sources

RTAB-Map 官方主页（IntRoLab） (Chinese)