Abstract:
Unmanned aerial vehicles (UAVs), particularly quadrotors, have become indispensable in both military and civilian applications, offering autonomous capabilities that enhance operational efficiency. Among the many challenges faced during autonomous UAV missions, landing is one of the most critical tasks, requiring precise localization of the landing spot and smooth, controlled trajectories for safe and accurate landings. The quadro-tor landing problem has significant implications for real-world applications, such as delivery services, search and rescue operations, and military missions, where efficient and precise target tracking and landing are of critical importance. This research aims to address the problem of robust and precise landing of a quadrotor UAV onto a moving planar target. Given the superior tracking capabilities of vision-based control tech-niques, this work primarily focuses on developing dynamically feasible Image-BasedVisual Servoing (IBVS) schemes for quadrotors to land on arbitrarily moving targets. The thesis is structured around two main sub-problems: (1) the development of a dynamically feasible visual servoing velocity controller that enables the quadrotor to accurately track and land on a planar target with arbitrary linear and angular motion, and (2) the design of a trajectory tracking control system that allows the quadrotor torobustly follow the velocity reference derived from the IBVS controller while account-ing for external disturbances, such as the ground effect. This research presents several key contributions, including developing a generalized IBVS method for tracking targets with arbitrary translational motion with realistic target dynamics. This method is further extended into a feasible kinematic IBVS technique for quadrotors, with experimental validation demonstrating its effectiveness. In addition, the thesis discusses the design of a trajectory tracking controller for the quadrotor, considering partial knowledge o fdesired trajectories (with higher derivatives unknown), which serves as an initial step toward solving the second sub-problem. Further, the thesis focuses on integrating a vision-based dynamic controller for landing on a moving target under state and control input-dependent uncertainties. Simulations considering realistic environments validate the effectiveness of the proposed controllers, showcasing their ability to achieve precise landing performance despite considering dynamic targets.
