FSC Autopilot

The FSC Lab's autopilot package for controlling multirotor in OFFBOARD mode via mavros.

Compared to previous controller packages, our controller is:

• Robust: Our multirotor controller ensures safety and reasonable performance in spite of

  1. Motor thrust mismatch
  2. Vehicle weight mismatch
  3. Strong winds and aerodynamic effects

  thanks to a UDE (uncertainty and disturbance estimator)-based design

• Cross-Platform: Our framework works on drones running either Ardupilot or PX4

• Tested: Our controller has been deployed on both a lightweight DJI F450 quadrotor (1.7kg) AND a heavier T15 octorotor

Dependencies

We depend on cmake 3.24+. Follow these instructions here to install the latest CMake. Remember, Recent versions of CMake are available absolutely everywhere.

We depend on Eigen>=3.7 and a few standard ROS packages

sudo apt-get install libeigen3-dev \
  ros-$ROS1_DISTRO_NAME-mavros \
  ros-$ROS1_DISTRO_NAME-mavros-msgs \
  ros-$ROS1_DISTRO_NAME-geometry-msgs \
  ros-$ROS1_DISTRO_NAME-nav-msgs \
  ros-$ROS1_DISTRO_NAME-sensor-msgs \
  ros-$ROS1_DISTRO_NAME-tf2-eigen

where $ROS1_DISTRO_NAME is to be replaced with your ROS distribution name, e.g. noetic

Building

Simply run

catkin build

Simulation

PX4 SITL

PX4's gazebo classic with ROS1 simulation is by far the easiest for getting started. Once you have set up the simulation environment per PX4 documentation, you can launch the simulation (alongside mavros) with

roslaunch px4 mavros_posix_sitl.launch

Once the simulator is up and running, you can start our controller with

roslaunch fsc_autopilot_ros autopilot.launch builtin_param_file:=gazebo_iris_params.json

🔎 You can inspect these default parameters in params/gazebo_iris_params.json.

---

Alternative Positioning

Gazebo lets you directly access ground truth pose and velocity of the quadrotor via the P3D plugin. To let our autopilot receive ground truth feedback (instead of from /mavros/local_position/odom by default), you should remap the ground truth topic to /fsc_autopilot/position_controller/feedback.

💡 If you pass the feedback_topic:=<your topic name> argument to autopilot.launch, the launch file handles the remapping for you

---

Engage the controller by sending setpoints to the controller on the topic /position_controller/target In the simplest case, you can send plain position/yaw setpoints on the command line

rostopic pub /fsc_autopilot/position_controller/reference fsc_autopilot_msgs/PositionControllerReference "
position: {x: 0.0, y: 0.0, z: 0.0}
yaw: 0.0
"

🔎 You can inspect the message definition at fsc_autopilot_ros/fsc_autopilot_msgs/msg/PositionControllerReference.msg in order to write more involved (e.g. path planning) programs sending out these setpoints

Ardupilot SITL

We use the baseline Ardupilot SITL without Gazebo.

❓ Ardupilot currently supports Gazebo Garden onwards, which cannot be installed alongside ROS1 on Ubuntu 20.04; Its support for Gazebo classic depends on a community plugin.

Once you have set up the simulation environment per Ardupilot documentation, you can launch the simulation with

sim_vehicle.py -v copter -w  --mavproxy-args="--streamrate=100" --map --console

⚠️ The --mavproxy-args="--streamrate=100" argument is absolutely necessary to ensure a healthy IMU publishing rate

Then launch mavros and let it connect with the simulation with

roslaunch mavros apm.launch fcu_url:="udp://127.0.0.1:14550@"

Once the simulator and mavros are up and running, you can start our controller with

roslaunch fsc_autopilot_ros autopilot.launch builtin_param_file:=apm_sitl_params.json