rsasaki0109/lidar_localization_ros2

What it solves

This project provides a map-based 3D LiDAR localization system for ROS 2 and Nav2, allowing robots to determine their precise position and orientation within a pre-existing 3D point-cloud map (such as .pcd or .ply files).

How it works

The system uses NDT (Normal Distributions Transform) and GICP (Generalized Iterative Closest Point) algorithms to align current LiDAR scans with a prior map. It incorporates odometry and IMU data for prediction, performs continuous-time deskewing of scans to remove motion distortion, and supports guarded global initialization to help the robot find its starting position without manual guessing.

Who it’s for

It is designed for developers and roboticists working with ROS 2 (specifically Jazzy and Humble) who need reliable 3D localization for mobile robots, including those using Livox MID-360 sensors.

Highlights

• Support for NDT/GICP localization against .pcd and .ply maps.
• Integration with Nav2 and standalone operation modes.
• Continuous-time scan deskewing to maintain accuracy during motion.
• Guarded global initialization using occupancy maps or reference CSVs.
• Pre-configured launch files for specific hardware like the Livox MID-360.