GrAco, is a dataset that includes both ground and aerial views. The main characteristics of our dataset are as follows:

• Multimodal-Heterogeneous data: An available and completely public dataset for C-SLAM research. It includes multi-modal data (inertial measurement unit (IMU), cameras, Light Detection and Ranging (LiDAR), and Global Positioning System (GPS)) in large-scale urban scenes collected by a fleet of ground and aerial vehicles.
• High precision: sensors are synchronized with millisecond-level and well calibrated. Centimeter-level ground truth localization obtained from RTK GNSS is provided.
• Loop closures between robots: Encounters among robots are designed during data collection, providing useful scenarios to the C-SLAM research.

Here is the download page provided for convenience. For more details, please visit our dataset website: https://sites.google.com/view/graco-dataset.

Contributors

Yilin Zhu, Yang Kong, Yingrui Jie, Shiyou Xu and Hui Cheng from SYSU RAPID Lab.

If our work has helped you, please cite:

@article{DBLP:journals/ral/ZhuKJXC23,
  author    = {Yilin Zhu and
               Yang Kong and
               Yingrui Jie and
               Shiyou Xu and
               Hui Cheng},
  title     = {GRACO: A Multimodal Dataset for Ground and Aerial Cooperative Localization and Mapping},
  journal   = {{IEEE} Robotics Autom. Lett.},
  volume    = {8},
  number    = {2},
  pages     = {966--973},
  year      = {2023}
}

1. Data format
2. Download
   • Ground sequence
   • Aerial sequence

We provide data in rosbag file format (ROS1 and ROS2 format), and sift out six ground sequences and eight aerialsequences. In addition, we provide a relatively small sample sequence for ground and aerial respectively. Here are topics (ROS1 format) that each bag has.

We provided a total of 6 ground sequences and 8 air sequences. Acquisition equipment and sequences are shown in the corresponding pictures. The duration and length of each sequence are shown below.

The suffix in the name represents the flight altitude.