Autonomous precision landing for UAVs

In modern industry, automated technologies are particularly useful in simplifying several applications. This benefit is emphasized for performing dangerous and difficult-to-handle operations. Especially, Unmanned Aerial Vehicles (UAVs) are developed for operations without a pilot on board and can be used in contexts where human intervention is not directly possible. These types of vehicles have several features including safety, rapid data extraction, and accessibility to remote locations. These also enable new application opportunities in both indoor and outdoor environments (such as emergency situations following a catastrophic event, search and rescue missions, or even aerospace applications). All these case studies share the requirement of using fully autonomous aerial vehicles in such a way as to, partially or totally, eliminate human intervention. In addition, there are many benefits such as speed, safety, efficiency, low cost, quality, and accuracy as a result of the ever-increasing growth in market demand in recent decades. The present master thesis work aims to realize an architecture of an octocopter drone and its autonomous behaviour in the FIXIT project. In particular, regarding the precision landing of the UAV on board the rover. The FIXIT project is being developed by the Competence Industry of Manufacturing 4.0, located in Turin, Italy. The drone is an unmanned system capable of flying in industrial environments, both outdoors and indoors by managing specific missions, collecting and processing data, performing a stable and autonomous flight in any scenario thanks to an integrated obstacle avoidance algorithm, and, at the end of the task, perform a precision landing on-board an autonomous ground robot. Among the many solutions currently available to perform a precision landing, some use optical guidance systems, predictive control model methods, computer vision-based recognition systems, and systems using GNSS/RTK or UWB positioning. The solutions analysed, have considerable errors in the landing phase that would cause a failed landing. In order to provide a precision landing with an adequate accuracy, a combined approach of the above solutions, considerably improves the accuracy. At the preliminary stage, the adopted solution uses two localization systems having at most 10 cm of error, which is considered an optimal threshold for the project. In the outdoor environment, GNSS positioning with a Real Time Kinematics system was used, which allows a correction factor to be sent from the ground station to the moving aircraft, which will improve the data received from the satellite; in the indoor environment, on the other hand, UWB technology with a tag and a POZYX anchor system was chosen, which allows its localization in the absence of signal from GNSS satellites. Once the accurate location was obtained, a computer-vision based system was developed, which uses a camera to identify a unique marker placed on board the rover. Following the identification of the marker, the landing phase will be carried out by keeping the marker in the field of view of the on-board camera; this will allow it to remain on the landing site during the deployment even in the presence of external disturbances that could displace it from the final target.