This is a ROS 2 package for automated generating of large UAV datasets. It currently supports generation of synchronized 3D position trajectories, GPS coordinates, IMU readings, RGB and depth images.
NOTE This work is submitted for publication to ICUAS 2024. Dataset will be released upon acceptance.
- ROS 2 humble + Gazebo
garden
- PX4 Atuopilot
The simulation deveopment environment is available in a self-contained docker image. First follow the instructions at d2dtracker_sim_docker
-
Combile the workspace using
colcon build
-
Source the workspace
source install/setup.bash
-
Launch Simulation: In your first terminal, initiate the simulation by running:
ros2 launch uav_dataset_generation_ros random_trajectories.launch.py
Upon execution, Gazebo should display a quadcopter.
-
Run QGroundControl: Open a second terminal and launch QGroundControl:
./QGroundControl.AppImage
This action should showcase the quadcopter within QGroundControl.
Once the setup is complete, initiate takeoff and switch to offboard mode to observe the drone execute random trajectories within the simulation environment.
If you use this repository, please cite
@software{Abdelkader_uav_dataset_generation_ros_2024,
author = {Abdelkader, Mohamed and Gabr, Khaled},
month = jan,
title = {{uav_dataset_generation_ros}},
url = {https://github.com/mzahana/uav_dataset_generation_ros},
version = {1.0.0},
year = {2024}
}