LT-mapper: A Modular Framework for LiDAR-based Lifelong Mapping

LT-mapper is an open-source, modular framework designed for LiDAR-based lifelong mapping in dynamic environments. It enables robots to maintain accurate 3D maps over long time scales, addressing challenges such as:

• Multi-session trajectory alignment (LT-SLAM)
• High/low dynamic object filtering and change detection (LT-removert)
• Efficient map management and change composition (LT-map)

The framework supports real-world scenarios, handling ephemeral and persistent changes across multiple sessions without requiring perfect initial alignment.

For more deatils, please refer to our paper and video.

Features

• Multi-Session SLAM (LT-SLAM): Align multiple sessions in a shared coordinate system using anchor-node-based pose graph optimization.
• Dynamic Object Removal (LT-removert): Efficiently filter high dynamic points and detect low-dynamic changes between sessions.
• Map Management (LT-map): Maintain live maps, generate meta-maps, and efficiently compose changes using delta maps.
• Seamless Modular Integration: Each module can run seperately via file-based in/out protocol.

Getting Started

Prerequisites

• ROS (tested with Melodic and Noetic)

sudo apt install -y ros-distro-navigation
sudo apt install -y ros-distro-robot-localization
sudo apt install -y ros-distro-robot-state-publisher

• GTSAM

sudo add-apt-repository ppa:borglab/gtsam-release-4.0
sudo apt install -y libgtsam-dev libgtsam-unstable-dev

Installation

Clone the repository

cd ~/catkin_ws/src
git clone https://github.com/gisbi-kim/lt-mapper

Build the package

cd ..
catkin build ltslam removert

Using Docker

cd ~/catkin_ws/src/lt-mapper/docker
bash build.sh
bash run.sh

or you can pull the pre-built image from Docker Hub.

docker pull dongjae0107/lt-mapper:latest
bash run.sh

How to use it

0. Single-session Data Generation

To run the LT-mapper, you first need to generate a set of session data containing the following files:

• Keyframe Point Clouds
• Keyframe Scan Context Descriptors (SCDs)
• Pose Graph File

You can generate these session data files using the saver tool provided with SC-LIO-SAM. Alternatively, the saver is also available in SC-A-LOAM and FAST_LIO_SLAM.

1. LT-SLAM

LT-SLAM module aligns trajectories across multiple sessions. For a step-by-step guide, check out this tutorial video.

# modify paths in ltslam/config/params.yaml before you run
roslaunch ltslam run.launch

2. LT-removert & LT-map

LT-removert & LT-map module utilize Removert to clean single-session data and detect changes between two sessions.

# modify paths in ltremovert/config/params_ltmapper.yaml before you run
roslaunch removert run_ltmapper.launch

Outputs

Once the process is complete, following outputs will be generated:

Example: LT-mapper on ParkingLot dataset

You can download the ParkingLot dataset from this link. To replay the data, you may need a file player.

ParkingLot dataset includes 6 sequences captured over three days at the same spatial site but with varying initial poses.

Step 1: Run LT-SLAM

Update paths in ltslam/config/param.yaml and execute LT-SLAM.

LT-SLAM will automatically align multiple sessions within a shared coordinate system.

Step 2: Run LT-Removert & LT-Map

Update paths in ltremovert/config/param_ltmapper.yaml and execute LT-removert & LT-map.

After running LT-Removert & LT-Map, you will get cleaned session data and detected changes between sessions.

Citation

@inproceedings{kim2022lt,
  title={Lt-mapper: A modular framework for lidar-based lifelong mapping},
  author={Kim, Giseop and Kim, Ayoung},
  booktitle={2022 International Conference on Robotics and Automation (ICRA)},
  pages={7995--8002},
  year={2022},
  organization={IEEE}
}

Contact

• Maintained by Giseop Kim and please contact the author via gsk@dgist.ac.kr

Acknowledgement

• Big thanks to Dongjae Lee for contributions in refactoring, dockerization, documentation, and various verifications.

Contributors