UAV net landing system for marine operations

Abstract

This thesis presents an automatic landing system designed for a fixed-wing unmanned aerial vehicle (UAV) to land in a net placed on ta ship. The UAV used is a Skywalker X8, a cheap and durable airframe, with a Pixhawk as a low-level autopilot while the high-level autopilot is written with the LSTS toolchain developed at The Underwater Systems and Technology Laboratory in Portugal. The high-level autopilot calculates desired speed and attitude with decoupled lateral and longitudinal line-of-sight (LOS) guidance controllers, while the low-level autopilot calculates desired elevon angles and throttle based on the desired attitude and speed from the high-level controller. The main objective was to design and develop a way of predicting the ship motion in seaway ensure successful landing attempts. Analysis of ship motion in seaway was conducted which resulted in that a version of artificial neural networks called long short-term memory networks (LSTM) was used to predict the separate motions of the ship. Software-in-the-loop (SITL) demonstrated successful performance of the automatic landing system and the LSTM showed good results when predicting the ship motion.