• autonomous landing
• autopilot
• computer vision
• fiducial markers
• guidance systems
• indoor navigation
• Kalman filter
• motion capture
• offboard control
• optical flow
• unmanned aerial vehicle
• visual servoing

This work of Thesis is concerned with the design, implementation and test of a precision autonomous landing system for a multirotor. It is involved in the Dromosplan European project, which is aimed at introducing swarm of drones in autonomous inspection, monitoring and surveillance of steel plants. In particular, the focus of this work is to make the drone capable of going to recharge autonomously on a landing pad of limited size. At first, the requirements of the application are analyzed, the methods and the equipment considered more suitable are selected by evaluating the state of the art and the available technology, and then implemented on a development aerial vehicle. The images from a bottom camera are processed to detect a set of markers placed on the landing pad to estimate the position and orientation with respect to the platform. A distance sensor and an optical flow device are integrated onboard to measure the distance and the horizontal velocity with respect to the ground, respectively. All data from the onboard sensors are fused by an extended and delayed Kalman filter to obtain precise and robust estimates of the vehicle state. Furthermore, a ground map of markers is employed for indoor navigation. Proportional Navigation is implemented to guide the drone in the approach phase to the platform, while a more aggressive PID control is applied in the final phase, with a smooth transition between the two. A safety check is also performed during the descend to ensure a reliable landing. Finally, indoor flight tests are repeatedly carried out to properly tune the Kalman filter, and then flight data are compared with the measures from the Vicon motion capture system, considered as ground truth.