Landing a Multirotor on a Vessel at Sea

Abstract

Operating unmanned aircraft from vessels in the open ocean is demanding both on the personnel involved in the operation and on the aircraft. The work presented here aims to alleviate the workload on pilots and other personnel by introducing a higher level of automation in the landing phase of a flight. This will help create value for all parts of the supply chain, from product developers through service providers, to the wind farm operators. It also has the potential to increase safety during operations and lower the demand for pilots with exceptional skills. We look at key parts of a system needed to perform vessel-based operations, with the emphasis on machine vision components employed in the task of detecting fiducial markers in order to estimate the position of the landing pad. We set up an extensive and complex simulation environment that lets us capture images resembling those from our chosen cameras. Using the images captured, we can estimate the position of ArUco markers in the frames. The thesis investigates the method of model predictive control to create trajectories towards the markers, and we do a brief comparison with a trivial solution. The simulator environment uses open source components and is designed to be close to reality with waves, a vessel, and an autopilot system, which we will use in the research project this thesis is a part of.