gaussian-rio-cpp

4D Radar-Inertial Odometry implementation in C++, based on Gaussian modeling and multi-hypothesis scan matching

[Preprint (arXiv)]

Under Construction

Requirements

• CUDA Toolkit (including nvcc)
• ROS (optional). We target the following versions:
  • ROS 1 One (currently needed for upo_gaussian_tester)
  • ROS 2 Humble
• small_gicp
• Ceres (system package)
• gaussian_server.py from gaussian-rio

For evaluating NTU4DRadLM and Snail-Radar trajectories:

• numpy
• evo

The generated trajectories used to calculate the metrics in our paper can be found in our website.

Packages

This repository contains two packages:

• upo_gaussians: A library providing the actual implementation of the 4D Radar-Inertial Odometry. It provides multiple scan matching algorithms, including Gaussians, NDT, GICP and VGICP. It can be used with or without ROS.
• upo_gaussian_tester: A ROS package providing a run_offline tool that allows for running the odometry directly on existing datasets (NTU4DRadLM, Snail-Radar)

In order to use the Gaussian odometry, it is currently necessary to run the gaussian_server.py script from our previous implementation, which is in charge of generating Gaussian models from radar point clouds. A native optimized implementation of Gaussian fitting is planned to be added in the future, which will remove this requirement.

Scripts

This repository contains the following scripts (in the scripts/ folder):

• run-ntu4dradlm.sh: Runs the odometry on four sequences of the NTU4DRadLM dataset.
• run-snail-radar.sh: Runs the odometry on five sequences of the Snail-Radar dataset.
• evaluate_ntu4dradlm.py: Generates evaluation metrics for NTU4DRadLM.
• evaluate_snailradar.py: Generates evaluation metrics for Snail-Radar.

Before running the odometry, it is required to adjust the path to the datasets in their respective .yaml files. Likewise, it is necessary to adjust the DATASET_DIR variable in the evaluation scripts.

Reference

@misc{gaussian4drio,
	author = {Fernando Amodeo and Luis Merino and Fernando Caballero},
	title = {4D Radar-Inertial Odometry based on Gaussian Modeling and Multi-Hypothesis Scan Matching},
	year = {2024},
	eprint = {arXiv:2412.13639},
}

Acknowledgements

This work was partially supported by the following grants: 1) INSERTION PID2021-127648OB-C31, and 2) NORDIC TED2021-132476B-I00 projects, funded by MCIN/AEI/ 10.13039/501100011033 and the "European Union NextGenerationEU / PRTR".