MOLA packages

MOLA’s modular design means a SLAM or localization problem is split into the smallest possible reusable pieces, each shipped as one colcon package. The ones that implement a mola_kernel interface (data sources, front ends, state estimators, map servers, visualizers) are modules: they are named in a YAML configuration file and loaded at runtime, without recompiling anything. The rest are the supporting cast, linked as ordinary libraries or used as build-time dependencies: message definitions, CLI tools and metapackages.

This page is the inventory: what each package is for, and where its sources live. For the tasks these packages are combined to solve, see solutions; for the C++ classes inside them, see the API reference.

Note

Each package carries its own license (GPL-3.0 for core algorithms, BSD-3-Clause for utilities, bridges and demos). Check the <license> tag of its package.xml.

Core framework

Everything else builds on these. They live in the MOLAorg/mola repository.

Package	Purpose
mola_kernel	The fundamental C++ virtual interfaces and data types every other module derives from (`RawDataSourceBase`, `FrontEndBase`, `NavStateFilter`, `MapServer`, `VizInterface`, …).
mola_launcher	Loads modules from a YAML configuration and runs their life cycle. Provides mola-cli, the main entry point.
mola_yaml	YAML parsing with MOLA’s extensions: variable expansion, includes, imports and formulas.
mola_msgs	ROS 2 messages, services and actions used by the other packages.

Input: datasets and sensors

Except where noted, each of these implements RawDataSourceBase, so any of them can feed any front end. See sensors and datasets for how to choose one.

Package	Purpose
mola_input_rosbag2	ROS 2 bag files (`.mcap`, sqlite3), including `/tf` resolution.
mola_input_rosbag1	ROS 1 `.bag` files, with no ROS 1 installation required.
mola_input_rawlog	MRPT `.rawlog` datasets.
mola_input_lidar_bin_dataset	Directories of `.bin` LiDAR scans in the KITTI binary layout.
mola_input_video	Live or offline video sources, via OpenCV.
mola_input_ouster	Ouster LiDARs through the native Ouster SDK: direct sensor connection and PCAP replay, with no ROS middleware in between.
mola_academic_datasets	Metapackage, not a module itself; it bundles the readers for public benchmark datasets, each of which is one: `mola_input_kitti_dataset`, `mola_input_kitti360_dataset`, `mola_input_mulran_dataset`, `mola_input_euroc_dataset`, `mola_input_paris_luco_dataset`, plus `kitti_metrics_eval`.
mola_bridge_ros2	Bidirectional ROS 2 ↔ MOLA bridge: subscribes to live topics as a data source, and publishes MOLA’s outputs back as topics and `/tf`.

Odometry, SLAM and state estimation

Package	Purpose
mola_lidar_odometry	3D/2D LiDAR and LiDAR-inertial odometry, the main MOLA front end.
mola_mapper	Central 3D SLAM map: online fusion of LIO/VIO/IMU/GNSS/wheels with loop closure and geo-referencing.
mola_state_estimation	State estimators (`Simple` and `Smoother`), plus `mola_georeferencing` and `mola_gtsam_factors`.
mola_imu_preintegration	IMU preintegration and gravity estimation used by the LIO paths.
mola_sm_loop_closure	Offline loop-closure post-processing of a simple-map, as a library and a CLI tool.
mola_relocalization	Global localization and pose estimation under large initial uncertainty.

Maps, tools and visualization

Package	Purpose
mola_metric_maps	The map classes: `KeyframePointCloudMap`, `IncrementalPointCloud`, `NDTMap`, `SparseVoxelPointCloud`, super-resolution occupancy grids, and the `.mm` CLI tools around them.
mp2p_icp	ICP pipelines, point-cloud filters and map tools.
mola_pose_list	Spatially-searchable pose lists (keyframe density bookkeeping).
mola_traj_tools	CLI tools to manipulate trajectory files, complementing `evo`.
mola_gnss_to_markers	Publishes GNSS datums as RViz/FoxGlove ellipsoid markers on a geo-referenced map.
mola_viz / mola_viz_imgui	The two interchangeable GUI backends (nanogui and Dear ImGui). See choosing a visualizer.
mola_demos	Example `mola-cli` launch files and ROS 2 launch files.