How ariarobotics/robotic-mapping works
1. This repository is a course monorepo, not a single application. The top level is split into:
- course-wide documentation and lecture PDFs,
- reusable resource/docs,
- several self-contained C++ example projects under code/cpp.
2. The lectures/ and resources/ directories are the instructional layer. They contain the curriculum PDFs, setup guides, dataset lists, and tooling references. These files support the code projects but do not participate in runtime execution.
3. The data/ directory at the repo root is a shared sample dataset area. It currently contains a KITTI sequence (KITTI07) used by the visual odometry examples. Individual subprojects also carry their own local data/ folders, so data is intentionally scoped per project.
4. code/cpp/competition is a standalone CMake-based C++ exercise centered on synthetic image data. Its bundled assets include camera calibration, trajectories, image sequences, and timing data, which suggests a pipeline for evaluating mapping/localization performance against synthetic ground truth.
5. code/cpp/features is another isolated C++ project with its own src/ and data/ folders. Based on the name and course context, it is the front-end feature extraction/matching exercise, likely meant to consume images and produce feature-level outputs for later SLAM stages.
6. code/cpp/visual-odometry is a self-contained visual odometry module. Its data layout mirrors a classic VO pipeline: calibrated image sequence plus timestamps, with source files vo_epipolar.cpp and vo_epipolar_noviz.cpp indicating a core epipolar geometry pipeline and a non-visualization variant.
7. code/cpp/lidar-odometry is a separate lidar front-end example. It uses a sequence of PLY point clouds (cloud_0.ply through cloud_4.ply) and a single main.cpp, indicating a small odometry pipeline focused on scan-to-scan pose estimation from point clouds.
8. code/cpp/lio-slam is the most complete SLAM subsystem and is organized as a ROS2 workspace with multiple packages:
- scanmatcher for lidar scan matching and front-end processing,
- graph_based_slam for back-end optimization and loop-closure style graph refinement,
- lidarslam_msgs for custom message definitions,
- Thirdparty/ndt_omp_ros2 for an embedded registration library.
9. The main data flow in lio-slam is front-end to back-end:
- sensor/scan data is processed in scanmatcher (image_projection.cpp, imu_preintegration.cpp, scanmatcher_component.cpp),
- outputs are turned into ROS messages using lidarslam_msgs,
- those outputs feed graph_based_slam for global consistency and map refinement.
The presence of launch/ and rviz/ folders shows the system is intended to be run and inspected as a live ROS2 graph.
10. scanmatcher contains build, install, and log artifacts inside its tree, which means the repository includes a snapshot of a built ROS2 workspace, not just source. For architecture diagrams, those generated directories should be treated as non-source outputs, not separate subsystems.
11. Thirdparty/ndt_omp_ros2 is an embedded external dependency rather than course-authored code. It provides optimized NDT/GICP/OpenMP point-cloud registration implementations and ships with its own CMake, headers, sources, and test data (.pcd files). This is the main external algorithmic dependency in the SLAM stack.
12. Build/runtime tooling is split by subsystem:
- the small examples use plain CMake,
- the SLAM system uses ROS2 package manifests, launch files, YAML params, and RViz configs,
- .devcontainer/ folders appear in several projects for reproducible development environments.
This means the repo supports both simple native builds and ROS2 workspace-style builds.
13. Across the repository, the main architectural boundary is between:
- teaching material and datasets,
- standalone algorithm exercises,
- the integrated ROS2 SLAM stack.
The course examples are intentionally modular so each front-end method can be studied independently before being combined in the full LIO/SLAM pipeline.
Main entry points: README.md Course repo · lio-slam LIO-SLAM · visual-odometry Visual odometry · scanmatcher Scanmatcher
Components 18
Instructional
Course-wide lectures and setup resources.
Shared Data
Repository-level datasets used across examples.
- KITTI07 · Repository-level sample KITTI sequence used by VO examples.
data/KITTI07
Standalone C++
Self-contained CMake exercises and front-end examples.
- Competition · Synthetic-image mapping/localization exercise with bundled assets.
code/cpp/competition - Competition core · Main executable for synthetic competition data processing.
code/cpp/competition/src/competition.cpp - Features · Standalone feature extraction and matching exercise.
code/cpp/features - Feature code · Feature-processing implementation for the course exercise.
code/cpp/features/src/features.cpp - Visual odometry · Self-contained epipolar visual odometry example.
code/cpp/visual-odometry - Epipolar VO · Core epipolar geometry pipeline for monocular VO.
code/cpp/visual-odometry/src/vo_epipolar.cpp - VO no-viz · Headless variant of the epipolar VO pipeline.
code/cpp/visual-odometry/src/vo_epipolar_noviz.cpp - Lidar odometry · Small point-cloud odometry example over a short sequence.
code/cpp/lidar-odometry - Lidar main · Single executable driving the lidar odometry pipeline.
code/cpp/lidar-odometry/src/main.cpp
ROS2 LIO-SLAM
Integrated ROS2 workspace for lidar-inertial SLAM.
- LIO-SLAM · Integrated lidar-inertial SLAM system with live launch and visualization.
code/cpp/lio-slam - Scanmatcher · ROS2 package for scan matching, projection, and IMU preintegration.
code/cpp/lio-slam/scanmatcher - Graph SLAM · ROS2 package for global optimization and map refinement.
code/cpp/lio-slam/graph_based_slam - Lidar msgs · Custom message definitions connecting SLAM components.
code/cpp/lio-slam/lidarslam_msgs - NDT OMP · Embedded external dependency for optimized point-cloud registration.
code/cpp/lio-slam/Thirdparty/ndt_omp_ros2
Other components
- Course repo · Top-level repository overview and navigation.
README.md
Connections 21
- Course repo → Lectures: documents · Course overview points to instructional material.
- Course repo → Resources: references · Top-level repo guidance references setup resources.
- Course repo → Competition: includes · Repository contains standalone C++ exercises.
- Course repo → Features: includes · Repository contains standalone C++ exercises.
- Course repo → Visual odometry: includes · Repository contains standalone C++ exercises.
- Course repo → Lidar odometry: includes · Repository contains standalone C++ exercises.
- Course repo → LIO-SLAM: includes · Repository contains integrated ROS2 SLAM stack.
- Competition → Competition core: builds · CMake exercise centers on a single executable source.
- Features → Feature code: builds · Feature exercise centers on its main source file.
- Visual odometry → Epipolar VO: core · Primary visual odometry implementation.
- Visual odometry → VO no-viz: variant · Non-visualization variant of the same pipeline.
- Visual odometry → KITTI07: uses · Uses the shared KITTI sequence from the repo root.
- Lidar odometry → Lidar main: builds · Single-file odometry pipeline entry point.
- LIO-SLAM → Scanmatcher: front-end · Scan matching package handles sensor processing.
- LIO-SLAM → Graph SLAM: back-end · Graph optimization package handles global consistency.
- LIO-SLAM → Lidar msgs: messages · Custom ROS2 messages connect workspace components.
- LIO-SLAM → NDT OMP: depends on · Embedded third-party registration library.
- Scanmatcher → Lidar msgs: publishes · Front-end outputs are encoded as ROS messages.
- Scanmatcher → Graph SLAM: feeds · Front-end outputs flow into back-end optimization.
- Scanmatcher → NDT OMP: registers · Scan matching relies on optimized point-cloud registration.
- LIO-SLAM → KITTI07: shares data · Course dataset is reused across examples in the repo.