ReFeree-Radar-SLAM

[IEEE RA-L] This repository is the official code for ReFeree: Radar-Based Lightweight and Robust Localization Using Feature and Free space.

Hogyun Kim*, Byunghee Choi*, Euncheol Choi, Younggun Cho^†

(* represents equal contribution.)

Spatial AI and Robotics Lab (SPARO)

Organization

• This repository is using the structure of Navtech-Radar-SLAM. The odometry and pose-graph optimization part is Yeti-Odometry and iSAM2, respectively.
• The Radar Place Recognition parts of this repository are referee.hpp and referee.cpp, and matching part in odometry.cpp.

Preliminary

This repository built and complied in environment Ubuntu 20.04, ROS Noetic(desktop-full). The required apt install packges are below.

$ sudo apt-get install python3-catkin-tools
$ sudo apt-get install ros-noetic-gtsam

How to use

First, clone and build the repository in your ros workspace.

$ cd your_ros_workspace/src
$ git clone https://github.com/sparolab/ReFeree-Radar-SLAM.git
$ cd your_ros_workspace
$ catkin build nano_gicp && catkin build

Second, prepare dataset in following format:

(dataset_name)/  
├── polar/  
│ ├── (timestamp_start).png  
│ ├── ...  
│ └── (timestamp_start).png

The timstamp-named png radar image files should be located in (dataset_name)/polar directory. And revise seq_dir and save_directory parameters in yeti_radar_odometry.launch.

<param name="seq_dir" type="string" value="(path_of_your_dataset)/(dataset_name)/" />
...
<param name="save_directory" type="string" value="(path_of_your_results)/results/"/>

Third, launch the yeti_radar_odometry.launch.

$ source devel/setup.bash
$ roslaunch yeti yeti_radar_odometry.launch 

Special Thanks

• Yeti-Odometry
• Nano-GICP
• Navtech-Radar-SLAM

Citation

@article{kim2024referee,
  title={ReFeree: Radar-Based Lightweight and Robust Localization using Feature and Free space},
  author={Kim, Hogyun and Choi, Byunghee and Choi, Euncheol and Cho, Younggun},
  journal={IEEE Robotics and Automation Letters},
  year={2024},
  publisher={IEEE}
}

Contact

• Hogyun Kim ([email protected])
• Byunghee Choi ([email protected])