AgriLiRa4D: A Multi-Sensor UAV Dataset for Robust SLAM in Challenging Agricultural Fields

Detail link: https://github.com/zhan994/AgriLiRa4D

To facilitate data organization and access, AgriLiRa4D is structured into sequence data, ground-truth trajectories, and calibration files, as summarized below.

AgriLiRa4D/
├── data/
│   ├── flat/
│   │   └── NJFlatB01/
│   │       ├── README.md
│   │       └── metadata.json
│   │       └── lidar_frames.csv
│   │       └── radar_frames.csv
│   │       └── imu.csv
│   │       └── fins_rtk_pose_enu.csv
│   │       └── fins_rtk_pose_flu.csv
│   │       └── fins_rtk_llh.csv
│   │       └── lidar.zip
│   │       └── radar.zip
│   │       └── raw.bag
│   ├── hill/
│   └── terr/
├── gt_flu/
└── calib.yaml
└── imu.yaml

LiDAR and radar filenames use the original ROS header timestamp. Their ordered frame tables provide explicit indices and retain both header and bag timestamps.

ROS topic	Open-format output
/rslidar_points	lidar/<timestamp>.pcd and lidar_frames.csv
/radar_points	radar/<timestamp>.pcd and radar_frames.csv
/rslidar_imu_data	imu.csv
/aircraft_pose_enu	fins_rtk_pose_enu.csv
/aircraft_pose_flu	fins_rtk_pose_flu.csv
/aircraft_position_llh	fins_rtk_llh.csv

Citation

@misc{zhan2025agrilira4dmultisensoruavdataset,
      title={AgriLiRa4D: A Multi-Sensor UAV Dataset for Robust SLAM in Challenging Agricultural Fields}, 
      author={Zhihao Zhan and Yuhang Ming and Shaobin Li and Jie Yuan},
      year={2025},
      eprint={2512.01753},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2512.01753}, 
}