Design Of Autonomous Landing System For UAV Based On Visual Servoing On Mobile Platform

The autonomous landing of UAVs means landing a UAV on a certain platform safely using flight control system and navigation devices such as GPS and IMU.With the widespread use of UAVs in military and civilian applications,autonomous landing navigation of UAVs has become an urgent problem to be solved.In this paper,the quadrotor is taken as the research object,and the image-based visual servo control strategy is used to design the autonomous landing control system of the quadrotor.The main work of this thesis is listed as followed:(1)The physical structure and flight principle of the quadrotor were analyzed before its kinematics and dynamics were modeled.Based on the auto disturbance rejection control method,a self-stabilizing controller was designed for the height and attitude system of the quadrotor,which was used to help quadrotor automatically return to the equilibrium state when the quadrotor's flight state changed.Finally,simulation experiments were carried out by Simulink to verify the controller effect.(2)A composited landing pattern was designed base on AprilTag which allowed the quadrotor to capture effective visual guidance at different stages of the landing phase.At the same time,a two-level cascaded neural network based on MobileNet was designed to perform key point detection of the landing pattern.The designed cascading key point detection neural network can achieve higher detection accuracy while satisfying the real-time requirements of embedded devices.(3)An image-based visual servo controller was designed and the controller's robustness is enhanced by introducing the height information into the interaction matrix.In the meantime,a feedforward compensation method was used to solve the influence of the ground target motion,and an adaptive gain coefficient was used to make the reference speed output computed by the visual servo controller smoother.Finally,the PID controller was used to realize the tracking of the reference speed signal of the visual servo controller by the quadrotor.(4)The designed visual servo control system was realized on the robot operating system ROS.The model of the quadrotor and landing scene were built in the 3D simulation platform Gazebo,in which the autonomous landing experiment was carried out.At the same time,the autonomous landing experiment of the quadrotor drone was carried out in the actual landing scene.Both simulation and physical experiments verify the effectiveness of the entire autonomous landing system and meet the needs of the actual application scenario.