(A) vision-based automatic landing of fixed-wing UAVs

Abstract

This thesis is focused on a vision-based landing system for fixed-wing unmanned aerial vehicles (UAVs) using an inflated airbag. As the statistics show, the landing stage is identified as most accident-prone among the entire flight regimes due to the complicated dynamics, difficult control characteristics and its dependency on many external factors such as wind conditions. Therefore there have been many attempts to automate the landing process on conventional runways or external arresting mechanisms such as ropes or nets. In this thesis, a relatively simple yet highly effective vision-based landing method using an inflated airbag is proposed. A Blended Wing-body(BWB) UAV hardware is set up for experiments and the identified system model is used for designing an autopilot controller. The airbag has a dome shape so that the airplane can approach from any direction, unlike the net-based approach that may suffer from crosswind. The dome`s distinctive color and shape provide unmistakably strong yet passive visual cue for the onboard vision system, which runs a fast and robust vision algorithm for geolocation and visual servoing. In terminal landing phase, the vehicle is controlled only by visual servoing commands without resorting to INS. The proposed idea has been validated in a series of experiments and the experiment results show that the proposed method is a viable approach for field applications without the need of expensive aiding systems.