hdl_graph_slam, an open-source ROS package published by Kenji Koide of Japan's National Institute of Advanced Industrial Science and Technology (AIST), is drawing renewed attention. It performs real-time 6DOF SLAM (simultaneous localization and mapping) using 3D LiDAR and NDT scan matching.
Open-Source 3D LiDAR SLAM · BSD-2-Clause
hdl_graph_slam: The Classic 3D LiDAR Mapping Package Back in the Spotlight
Feeding 3D LiDAR point clouds into NDT scan matching plus Graph SLAM, it delivers real-time 6-DoF pose estimation and mapping — free, flexible, and still underpinning field robotics.
6-DoF
Real-time pose: position (xyz) + orientation
5
Constraint types combinable as graph edges
$0
BSD-2-Clause license, published on GitHub
The SLAM pipeline — errors corrected by graph optimization
Point Cloud
3D LiDAR frames in
→
Odometry
NDT / GICP matching
→
Loop Closure
Re-recognize places
→
Graph Optimize
Nodes + edges → PCD map
Flexible constraints (graph edges)
Beyond odometry and loop closure, diverse constraints can be layered in to tighten the solution.
Odometry
Loop closure
GPS
IMU acceleration / orientation
Floor plane (RANSAC)
Environment: ROS Melodic / Noetic · Sensors: Velodyne HDL-32e / VLP-16, RoboSense 16ch · Matching: FAST_GICP (recommended) or NDT_OMP · ndt_resolution 0.5–2.0m indoor, 2.0–10.0m outdoor
Praised for
Practical graph mapping of mobile scans without an IMU
High-quality maps achievable with good parameters
Floor-plane constraint effective in flat indoor spaces
Open questions
Tuning strongly influences the outcome
Some features not working with Ouster LiDAR
KITTI evaluation questions · requests for ROS2 support
The bigger point: while generative AI grabs headlines, classic SLAM implementations like this — alongside LeGO-LOAM, Cartographer, LIO-SAM and the newer glim — still underpin real field development in mobile robotics, autonomous driving and indoor mapping.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…