The main repository for ICUAS'24 UAV competition
If you're interested in publishing papers based on this repository, please cite the following publication:
@article{Markovic2023,
doi = {10.1007/s10846-023-01909-z},
url = {https://doi.org/10.1007/s10846-023-01909-z},
year = {2023},
month = jul,
publisher = {Springer Science and Business Media {LLC}},
volume = {108},
number = {3},
author = {Lovro Markovic and Frano Petric and Antun Ivanovic and Jurica Goricanec and Marko Car and Matko Orsag and Stjepan Bogdan},
title = {Towards A Standardized Aerial Platform: {ICUAS}'22 Firefighting Competition},
journal = {Journal of Intelligent {\&}amp$\mathsemicolon$ Robotic Systems}
}
Teams interested in the ICUAS'24 UAV Competition are encouraged to check the rulebook. Instructions on how to apply are in the rulebook.
You can either manually install the UAV software stack by following
uav_ros_simulation instructions or simply
use Docker instead.
To install Docker on your system execute the following command:
curl https://raw.githubusercontent.com/larics/uav_ros_simulation/main/installation/dependencies/docker.sh | bash
Check out CHANGELOG.md
for any new changes added to this project.
Feel free to use the Discussions tab to exchange ideas and ask questions.
Consider opening an Issue if you have troubles with the simulation setup.
NOTE - If the challenge is not set up correctly it is (probably) not your fault! Components are subject to some changes during the competition so most problems should be solved by updating packages. Try following the troubleshooting recommendation. If the problem persists please post an issue.
In case there are new changes to the challenge repository:
git pull origin main --rebase
catkin build
In case the Docker container simulation is not working correctly (e.g. an update happened):
git pull lmark1/uav_ros_simulation:[[DISTRO]]
In case the simulation inside the Docker container is (still) not working correctly:
./docker_build.sh --build-args "--no-cache --pull" --[[DISTRO]]
If you're building all the packages natively, navigate to the uav_ros_simulation
folder and do the following:
git pull origin main
./installation/install.sh
# Navigate to catkin workspace (default is uav_ws)
catkin build
Update all code of the competition repo and build as follows:
git pull origin main
catkin build
You can either manually build all the packages on your system using the catkin build
command.
Alternatively, to build the ICUAS2024 Competition solution image please execute the following command:
./docker_build.sh
Additional arguments:
--focal
- Build Docker image for Focal distro--focal-nogpu
- Build Docker image for Focal distro (no dedicated graphics card)--build-args
- Append additional Docker build arguments, e.g. --no-cache
To automatically start and setup the challenge navigate to startup/challenge
and run:
./start.sh
This should automatically setup and start the challenge, as well as run your code (see solution section in session.yml)
- Commands that run your challenge solution (rosrun, roslaunch etc.) should be placed in the
session.yml
file. - Software configuration specific to the challenge should be placed in the
custom_config
folder.
NOTE If you are unfamiliar with the Docker or Tmux commands please check out this quick-start guide.
NOTE If you choose to run the challenge inside the docker environment, please run the container first using:
./docker_run.sh
Additional arguments:
--focal
- Run Focal distro container--focal-nogpu
- Run Focal distro container (no dedicated graphics card)--run-args
- Append additional Docker run arguments, e.g. --rm
NOTE Keep in mind this will start a new container so any changes you make inside that container will be lost if you remove the container.
The idea of the container is to easily integrate your code with the challenge flight stack. To do so, please add your code directly to this ROS package since it is copied to the container. Furthermore, feel free to edit Dockerfile.focal
or Dockerfile.bionic
files to
get all the resources and build your solution.
UAV simulation template startup. Tmux session is running on the left side, with the Gazebo client positioned on the right. |
For your reference, we have set up trajectory planning using TOPP-RA, which you can use by publishing two topics:
tracker/input_pose
- Send a waypoint (PoseStamped) to TOPP-RA. TOPP-RA then interpolates trajectory between the current UAV pose and the target waypoint and sends trajectory points (MultiDOFJointTrajectoryPoint) to topicposition_hold/trajectory
with a given rate. The position controller of the UAV receives the trajectory point as a reference and commands the motors.tracker/input_trajectory
- Generate a trajectory using the given sampled path in the form of waypoints (MultiDOFJointTrajectory). TOPP-RA then interpolates trajectory from the current UAV pose to the first point of the trajectory and interpolates trajectory between sampled path points. Once the trajectory is interpolated, each trajectory point is sent as a reference to the position controller via theposition_hold/trajectory
topic
To control the UAV directly, and to publish the trajectory that you generated via your solution, you need to use the following topic:
position_hold/trajectory
- Publish a trajectory point directly to the UAV position control
Current position reference (the last one sent to the position controller of the UAV) can be obtained via carrot/pose
topic, while the current pose of the UAV (in simulation) is available at odometry
topic.
Configuration files are placed in the startup/challenge/custom_config
folder.
The goal of the UAV is to count the total number of fruits of a given plant. The list of beds that the UAV needs to visit will be provided on the topic plant_beds
within the UAV namespace. The message is a string (std_msgs/String
) and contains the name of the plant and beds that need to be visited (eg. Pepper 1 2 8 12 19
), separated by whitespace. The number of beds may vary in a given mission.
Plant beds layout. |
Plants are simulated via textures on flat surfaces. The fruits of one variety of plants have the same color and shape and are simulated as 3D objects partially visible from either one or both sides of the plant. The shapes and colors of fruits on different variety of plants are different. There are three plant varieties: a) Tomato (red), b) Pepper (yellow), and c) Eggplant (purple).
More details on the challenge can be found in the competition rulebook. After the UAV and the world is spawned, plants will also be spawned in beds. One plant bed can contain up to three plants. Each plant may have a different number of fruits (including zero). Some plant beds in the world may be empty, but the ones listed in the mission specification through the plants_beds
topic will have at least one plant.
UAV in the simulation arena for the ICUAS 2024 challenge. |
The setup in this repo reflects how the world for evaluation will be constructed. Teams are welcome to make a permanent world for testing and development since spawning of a large number of plants in Gazebo takes some time.
Remember that you should run your code alongside the existing setup through the session file. Use the existing solution section. You can use either existing waits to spawn your nodes or set your nodes to listen to the following topic:
challenge_started
- AfterTrue
is published on this topic the UAV is set up correctly.plant_beds
- After the array of points of interest is published on this topic the arena is set up correctly.
Make sure to wait for data on both topics to safely run your code.
You can change plant locations through the session file or spawn them from a separate terminal. Use position args to change the location of the plant.