Automatic Landing System of a Quadrotor UAV Using Visual Servoing

This paper presents a fully autonomous system for a quadrotor unmanned aerial vehicle (UAV) that employs the visual measurements to perform automatic landing task on a specific landing platform. Basically, there are two main control tasks discussed in this paper. The first task refers to the auto-land mission that tracks the platform horizontally and performs the vertical landing. It was accomplished by using the red blob tracking technique. The second task involves the robust motion tracking that activates the recovery action once the target is lost so that the vehicle is still able to hover steadily. It was realized by implementing the features accelerated from segment test (FAST) technique with the pyramidal Lucas-Kanade algorithm to detect the local feature in the image and compute the optical flow. From the visual results obtained, the position and velocity of the vehicle were estimated using a nested Kalman-based sensor fusion . The state estimation was validated in a series of experiments using a CNC milling machine. Lastly, the control architecture for automatic landing system was formed with the classical PID controller and the flight test proved the success of the proposed system.