Gazebo

Gazebo simulates drones in a 3D world with physics and sensors. Use it when a mission needs to see something, such as a camera, a lidar or ArUco markers, or when several drones share a world. Aerostack2 uses Gazebo Fortress on ROS 2 Humble and Gazebo Harmonic on ROS 2 Jazzy, which rosdep installs along with Aerostack2.

Two packages work together:

• as2_gazebo_assets builds the world from a configuration file, spawns the drones and their sensors, and bridges their topics to ROS 2.
• as2_platform_gazebo is the aerial platform. One runs per drone and sends its commands to Gazebo.

Launch

Start the simulation once, then one platform per drone:

ros2 launch as2_gazebo_assets launch_simulation.py \
  simulation_config_file:=config/world.yaml

ros2 launch as2_platform_gazebo platform_gazebo_launch.py \
  namespace:=drone0 \
  simulation_config_file:=config/world.yaml \
  platform_config_file:=config/config.yaml

The simulation opens the Gazebo window by default. With headless:=true it runs without one, and you can attach the window later with gz sim -g on Harmonic or ign gazebo -g on Fortress.

The world file

The simulation configuration file, in YAML or JSON, describes the world, the drones and what each drone carries:

world_name: "empty"
origin:
  latitude: 40.4405287
  longitude: -3.6898277
  altitude: 100.0
drones:
  - model_type: "quadrotor_base"
    model_name: "drone0"
    xyz: [0.0, 0.0, 0.3]
    flight_time: 60
    payload:
      - model_type: "gps"
        model_name: "gps"
      - model_type: "hd_camera"
        model_name: "camera"
        xyz: [0.1, 0.0, 0.0]

model_name becomes the drone's namespace. flight_time gives the drone a battery that lasts that many minutes, and leaving it out means no battery. origin places the world on the globe for GPS. An objects list can add more models to the world, such as gates or markers.

The code also accepts d435 and point_lidar, but d435 has no model yet and point_lidar is marked as not working.

Each sensor publishes under sensor_measurements/ with its name, so a camera called camera on drone0 publishes /drone0/sensor_measurements/camera/image_raw and camera_info. Every drone also publishes its IMU and ground truth, and its battery when it has one. There's no odometry topic, so use the ground truth estimator.

Control modes

By default the Gazebo platform takes velocity commands, with the yaw as a yaw rate, plus hover. That fits the PID speed controller. For controllers that send body rates, like differential flatness, launch the platform with the body rates control modes file that ships with the package:

ros2 launch as2_platform_gazebo platform_gazebo_launch.py \
  namespace:=drone0 \
  simulation_config_file:=config/world.yaml \
  control_modes_file:=$(ros2 pkg prefix as2_platform_gazebo)/share/as2_platform_gazebo/config/control_modes_acro.yaml

With the rest of the stack

Gazebo roots each drone's TF tree at the bare model name, drone0 rather than drone0/base_link. Set the base frame to an empty name so it resolves to the namespace:

/**:
  ros__parameters:
    use_sim_time: true
  state_estimator:
    ros__parameters:
      base_frame_id: ""
      plugin_name: "ground_truth"
      ground_truth:
        twist_sub_topic: "ground_truth/twist"

Gazebo publishes the simulation clock, so every node should run with use_sim_time on.

The project_gazebo repository has a complete setup with every component, a ground station and example missions for one or several drones.