Autoland project: Fixed-wing UAV Landing on a Fast Patrol Boat using Computer Vision

Abstract

The AUTOLAND project main objective was the development of solutions that can enable the autonomous landing of a fixed-wing Unmanned Aerial Vehicle (UAV) on a Fast Patrol Boat (FPB). Since we are operating in a military environment where jamming is possible, we have developed Computer Vision (CV) based systems without using any external sensor information. We have developed and tested two different CV approaches: (i) airborne, and (ii) ground-based. In the airborne approach, the UAV uses its camera and external markers to estimate the relative pose to the landing area and automatically calculates the needed landing trajectory. In the ground-based approach, a ground-based monocular vision system obtains the relative pose of the UAV concerning the camera reference frame using its Computer-Aided Design (CAD) model. Then, a Ground Control Station (GCS) calculates the landing trajectory and transmits it to the UAV. The obtained error was compatible with the automatic landing requirements. The future work will focus on real data acquisition to improve the developed algorithms.