Close Control of an AutonomousDrone for Physical ModuleReplacement on Mobile Platforms

Abstract

Fully autonomous drones are getting more and more reliable and arewilling to be used in an increasing amount of fields. The InternationalAerial Robotic Competition claims that repairing moving platformsin an autonomous way is an important one. This thesis will try toimprove the control of an autonomous drone close to a moving boat’smast. This includes the creation of an attitude controller through LQR,an Extended Kalman Filter to estimate the position of the movingtarget and some strategy about the movements around the mast. Afeed-forward and non-linear input to the LQR will be explored. Theresearch includes real-life validation and collision testing. The resultsof the LQR will be compared with an autopilot built-in PID controller.This thesis shows that the LQR could not get significantly better thanthe PIDs and that a good EKF tuning in terms of error makes thedrone jittering. For this kind of application, it is better to lose a bit ofaccuracy on the EKF side but make the flight smoother.