FORM

Fixed-Lag Odometry with Reparative Mapping

FORM is a LiDAR Odometry system that performs fixed-lag smoothing and sub-map reparations, all in real-time with minimal parameters.

---

Building

FORM has minimal dependencies. You will need:

• eigen3
• gtsam
• tbb
• tsl::robin_map (will be pulled from source if not found)

After that, building is as simple as:

mkdir build
cd build
cmake ..
make

Running Experiments from Source

All experiments are ran through evalio, our internal LiDAR-inertial odometry evaluation tool. Downloading the desired datasets is done as:

evalio dl newer_college_2020/*

FORM can then be installed as a python package (preferably in a virtual environment):

pip install -e .  # for pip usage
uv sync --verbose # for uv usage

Finally, experiments can be run as (with the venv activated if needed):

evalio run -M form -c config/25.10.03_full.yaml

Datasets can be commented out in the config file if they haven't been downloaded. Results are saved in evalio_results/25.10.03_full/ and stats can be viewed as:

evalio stats evalio_results/25.10.03_full/

Roadmap

• ROS Node
• Merge wrapper into upstream evalio

Citation

If you use FORM for any academic work, please cite:

@misc{potokar2025formfixedlagodometryreparative,
  title={FORM: Fixed-Lag Odometry with Reparative Mapping utilizing Rotating LiDAR Sensors}, 
  author={Easton R. Potokar and Taylor Pool and Daniel McGann and Michael Kaess},
  year={2025},
  eprint={2510.09966},
  archivePrefix={arXiv},
  primaryClass={cs.RO},
  url={https://arxiv.org/abs/2510.09966}, 
}