nanners.mp4
banana drone is an autonomous drone. it is built on the ROS2 framework, uses PX4 autopilot, and uses deep learning to do object detection. Inertial guidance allows it to fly patrol via waypoints.
first, you need to create a new Ros2 workspace. the, roughly follow these steps
install px4-autopilot
install drone_project_ws
install ros_gz_bridge
install Micro-XRCE-DDS-Agent
install darknet_ros_yolov4into drone_project_ws
install banana_drone
install gazebo
the PX4 x500 model doesn't come with a camera, so we can add one. Inside of the PX4-Autopilot repo apply the changes from
changes_for_camera.xml.
darknet_ros_yolo4into should be installed into the drone_project_ws/src directory.
These can be installed outside the drone_project_ws directory
Each step requires at least this much
cd drone_project_ws
. install/setup.bash
In 1 terminal run
ros2 run ros_gz_bridge parameter_bridge /camera@sensor_msgs/msg/Image@ignition.msgs.Image
In another terminal run
cd Micro-XRCE-DDS-Agent
MicroXRCEAgent udp4 -p 8888
In another terminal
ros2 launch darknet_ros darknet_ros.launch.py
In another terminal
make px4_sitl gz_x500
If it freezes and gazebo never starts, then hit ^Z, but it in the bg and run PX4_SYS_AUTOSTART=4001 PX4_GZ_MODEL=x500 ./build/px4_sitl_default/bin/px4
Finally, in another terminal
cd drone_prooject_ws/src
colcon build --packages-select drone_project
source install/setup.bash
ros2 launch drone_project drone_control_launch.py
You should re-run colcon everytime you make a change.
This code will allow the drone to fly in a pre-determined set of waypoints, like on a patrol. Non-banana objects of interest could be recorded and reported when the drone finishes its patrol. GPS is not needed.
The drone will hover and watch for an object, with the idea to chase a detected object. By keeping the object in the center aperature of the camera, the drone can follow or track an object without complicated algoirthms or GPS.