Trajectory planner Node in ROS2 to send trajectory "CSV file" to IRIS model in Gazebo step by step. Everything works inside a docker out of the box!!
We advise to use Docker to run both PX4 and the offboard application.
There is a convinient script that facilitate everything
curl -fsSL get.docker.com -o get-docker.sh
sudo sh get-docker.sh
We suggest to use docker as a non-root user, that way your build folder won't be owned by root after using docker. In a new terminal run
# Create docker group (may not be required)
sudo groupadd docker
# Add your user to the docker group.
sudo usermod -aG docker $USER
Now log out and in again before using docker!!
Open a terminal, go in the folder where you want to clone this repository and run the command
git clone https://github.com/alisrour97/Sensitivity_PX4_Experiments.git --recursive
Clone the firmware repository, in the same folder where you cloned this repository
git clone https://github.com/alisrour97/PX4-Autopilot.git --recursive
A forked repo of PX4 which contains another fork for gazebo-classic modules which was modified
This is the client needed to communicate with PX4 from ROS2. Clone version V2.4.0
git clone https://github.com/eProsima/Micro-XRCE-DDS-Agent.git
cd Micro-XRCE-DDS-Agent
mkdir build
cd build
cmake ..
make
Go into the sensitivity_experiment folder and run the script to create the container (N.B: you must be in the sensitivity_experiment folder)
cd Sensitivity_PX4_Experiments
source create_px4_container.sh
Now you should be inside the container, and if you run the command ls, you will see the two folders
To close the container just press CTRL-d. The container will be closed but not removed, so you can start it again whenever you need. If, instead, you need to remove the container for some reason, run the command
docker container rm px4_ros2
and repeat the step (5) to create it again.
Image can be found on the docker Hub
Now that everything is set up, you need to compile PX4 in order to run the simulation, start the uXRCE-DDS client and then start the controller in ROS2. Remember that to build PX4 and to run the controller you must be inside the docker container, otherwise you will miss the dependencies to do that (to start the uXRCE-DDS client you must not be in the container if you compiled it outside the container).
Considering that we will use the Gazebo GUI, in every terminal where you will start the container or open a terminal into it, first run the command
xhost +
If you closed the container, then start it with the command
docker start px4_ros2
docker attach px4_ros2
If instead you have already started the container in a terminal and want to open another terminal in the container, run the command
docker exec -u 0 -it px4_ros2 bash
In the container, go in the firmware folder and build it to start the simulation
cd PX4-Autopilot
make px4_sitl gazebo-classic
At the end of the building process you will see the Gazebo simulation opened.
Open a new terminal and go into the repository and start the client
cd Micro-XRCE-DDS-Agent/build
./MicroXRCEAgent udp4 -p 8888
Open a new terminal in the container and, only for the first time, build the ROS2 workspace
xhost +
docker exec -u 0 -it px4_ros2 bash
source ../../opt/ros/foxy/setup.bash
cd Sensitivity_PX4_Experiments
colcon build
At the end of the building process, or if you already have built before, run the commands
source install/setup.bash
ros2 run trajectory_publisher trajectory_publisher
You should see IRIS moving with feedforward trajectory of duration 5 seconds
cd Sensitivity_PX4_Experiments
A script to simulate a trajectory with 9 perturbed simulations
python3 -m simulate.py <trajectory_file_name> <trajectory_duration_in_seconds>
Make sure that you have permission to run simulate.py script by running
sudo chmod +x simulate.py
The log files are found when navigating to
cd ../PX4-Autopilot/build/px4_sitl_default/rootfs
Sometimes you have to run the following to get permission to plot logs in plotjuggler
sudo chown -R tesla:tesla *
my username is tesla
First you should be on the same wifi network as the drone WifiVicon, then you connect to the drone via ssh followed by username/id:
ssh nvidia@192.168.30.165
Make sure that the drone is recognized by Qualysis/Vicon on the Mocap desktop
The the next step would be to open the Mavlink shell on the px4 to enable communication
python3 PX4-Autopilot/Tools/mavlink_shell.py
Then you type the following:
micrortps_client start -t UART
mavlink stop-all
After that you can safely close that terminal
on your laptop you need the following package as to run atomatic script:
sudo apt install sshpass
Automatic script start_Acanthis.sh:
./start_Acanthis.sh
which does the following: connect to drone/ starting docker/ changing the directory to sensitivity experiments and sourcing the "setup.bash" which is translated in the following commands:
ssh nvidia@192.168.30.165
docker start sens_exp
docker exec -u 0 -it sens_exp
cd sens_ws/Sensitivity_PX4_Experiments
source install/setup.bash
Then you can start the MicroRTPS Agent to publish and subscribe to the drone nodes
micrortps_agent -t UART
In another terminal you also do the following as to make visp node "visp_odom" publish to the drone the position
./start_Acanthis.sh
ros2 run visp_odom visp_odom
Then in another terminal you can run the trajectory publisher with:
./start_Acanthis.sh
ros2 run trajectory_publisher trajectory_publisher --ros-args -p file_name:="your_file.csv"
To transfer CSV files, do the following:
scp directory_of_file/File.csv nvidia@192.168.30.165:~/sens_ws/sensitivity_experiment/src/trajectory_publisher/csv_file
To listen to vehicle different topics:
ros2 topic list
ros2 topic echo /fmu/vehicle_visual_odometry/in
ros2 topic echo /fmu/vehicle_odometry/out
ros2 topic echo /fmu/vehicle_control_mode/out
ros2 topic echo /fmu/vehicle_status/out
If in case you changed visp_odom node or trajectory publisher you have to build it again by:
colcon build --packages-select visp_odom --symlink-install
colcon build --packages-select trajectory_publisher
Ali Srour
Salvatore Marcellini
Tommaso Belvedere