Our main approach was to use Aruco markers and a PID controller to control the Drone. For that we used the libardrone library. The Aruco markers are detected with the opencv library.
We delivered a quite good race. Eventually we finished in place 5 out of 9. Our main issue were connection problems between the Drone and the Notebook, running the python script. This was especially severe in the TUM Maschinenbau building, because many people, many phones, and many Drones (each hosting its own WiFi hotspot).