robotics-upo / radar_experiments

Repository to store configuration files related to the smoke and radar experiments

Localization experiments.

This repository contains the source code and additional configuration files to reproduce the results of our work presented in the Robot 2022 international conference:

D. Alejo, R. Rey, J. A. Cobano, F. Caballero and L. Merino, "Data fusion of low-cost RADAR and LIDAR for reliable ground robotlocalization under low-visibility conditions". In D. Tardioli, V. Matellan, G. Heredia, M.F. Silva, and L. Marques, editors, ROBOT2022: Fifth Iberian Robotics Conference, Advances in Intelligent Systems and Computing, pp. 301–313, Springer International Publishing, 2023.

In this paper, we have designed and implemented a new method for fusing radar and lidar information. The method has been implemented mainly in the source file: src/sensor_fusion/lidar_radar_fusion_dbscan.cpp

The dataset used in this paper can be downloaded at:

Dataset web page

Dependencies

The utilities of this repository have been programmed on a Linux 18.04 distribution and ROS melodic. They have also been tested in Ubuntu 20.04 and ROS noetic. Please refer to the ROS wiki for details on installation and the creation of a catkin workspace.

• Cloud concatenator: https://github.com/robotics-upo/cloud_concatenator
• pcl_ros: sudo apt install ros-$ROS_DISTRO-pcl-ros
• pointcloud_to_laser_scan sudo apt install ros-$ROS_DISTRO-pointcloud-to-laserscan
• raposa_marker: https://github.com/robotics-upo/upo_markers
• odom_to_tf: https://github.com/robotics-upo/odom_to_tf

You can download the repository and execute ./scripts/install_fusion_dependencies.sh from the source directory of a catkin workspace. It will install APT packages and download the necessary repos to avoid doing it manually. In addition it asks you if you want to download the dataset.

Once you download our dataset into the folder "${HOME}/radar_fusion_bags", you can launch the experiment batch presented in the ICRA paper as follows (for ten repetitions):

rosrun radar_experiments experiment_batch.sh 1 10

It will run the fusion node and launch a rviz visualization window where you can select which radar scans you want to visualize.

Parameters for the implementation of P. Fritsche & B. Wagner 2017 "Modeling structure and aerosol concentration with fused radar and LiDAR data in environments with changing visibility" (IROS 2017)

Related to the algorithm:

• yaw_tolerance : Used to extract 2d virtual scans
• ransac_distance_threshold (for both lidar and radar)
• ransac_iterations (for both lidar and radar)
• min_ransac_pointcloud_size (for both lidar and radar)
• beta
• distance_threshold
• lidar_dev
• radar_dev
• lidar_max_range

Related to node:

• sychronization_margin
• time_tolerance

Acknowledgements

This work is partially supported by Programa Operativo FEDER Andalucia 2014-2020, Consejeria de Economía, Conocimiento y Universidades (DeepBot, PY20_00817) and by the Spanish Ministry of Science, Innovation and Universities (Insertion PID2021-127648OB-C31, MCIU/AEI/FEDER, UE).