FreeDOM is an online dynamic object removal framework for static map construction based on conservative free space estimation. With FreeDOM, you can create static voxel and pointcloud map in real-time, free from dynamic object interference.
Clone source code:
mkdir -p ./Freedom_ws/src
cd ./Freedom_ws/src
git clone https://github.com/LC-Robotics/FreeDOM.gitBuild:
cd ..
catkin buildRun experiments on the SemanticKITTI dataset:
source devel/setup.bash
roslaunch freedom run_freedom.launch
rosbag play DATASET.bagAll rosbags can be found here. To run experiments on other datasets (HeLiMOS, Indoor), modify the yaml file path in run_freedom.launch and play the corresponding rosbag.
-
When running KITTI seq.01 (high-speed scenario),
counts_to_freeshould be set to 3 to enable faster free space estimation. -
It is recommended to use
run_freedom_indoor.launchfor Indoor datasets for better visualization.
Save the generated static map:
rostopic pub /save_map std_msgs/Empty "{}" -1The static map will be saved by default at ./generated_pcd.
Generate ground truth:
roslaunch freedom ground_truth_generate.launch
rosbag play DATASET.bagSave the generated ground truth:
rostopic pub /save_map std_msgs/Empty "{}" -1The ground truth file will be saved by default at ./generated_pcd.
- Alternatively, you can directly download the ground truth file here.
Evaluation:
roslaunch freedom static_map_evaluate.launch-
Make sure the
ground_truth_pathandstatic_map_pathinstatic_map_evaluate.launchare correct. -
Note that we use a voxel size of 0.2m for outdoor datasets and 0.1m for indoor datasets, which can be modified in
ground_truth_generate.launchandstatic_map_evaluate.launch.
Please cite this paper if you find our work useful for your research.
@article{Li2025FreeDOM,
author={Li, Chen and Li, Wanlei and Liu, Wenhao and Shu, Yixiang and Lou, Yunjiang},
journal={IEEE Robotics and Automation Letters},
title={FreeDOM: Online Dynamic Object Removal Framework for Static Map Construction Based on Conservative Free Space Estimation},
year={2025},
volume={10},
number={6},
pages={5577-5584},
doi={10.1109/LRA.2025.3560881}
}