Vision-based autonomous landing of a model-scaled rotorcraft UAV

Abstract

Development of a reliable high-performance rotorcraft unmanned air vehicle(R-UAV) requires an accurate and practical model of the vehicle dynamics. This paper describes the analytical model driven by use of basic aerodynamics, providing the rationale for the derivation of the equations of motion as well as documenting the experimental methods for the estimation of a model-scaled unmanned helicopter. Hovering and autonomous landing simulation were carried out by using the nonlinear helicopter dynamics to predict how well the helicopter performs its missions. In addition to the helicopter modeling, the real flight test had been executed with a model-scaled helicopter. The main goal of this thesis is to develop and establish a comprehensive and practical methodology for vision-based autonomous landing of a R-UAV. To demonstrate this idea, it is essential to integrate the vehicle platform with the proper hardware and software in order to perform the desired autonomous landing. The whole system has been integrated with sensors such as a gyro, magnetometer and GPS, the Flight Control Computer (FCC) and the Vision Computer. For vision-based autonomous landing, we carried out the autonomous landing without vision-sensor to verify how the helicopter can land on certain area. After that, the altitude estimation is conducted by a single camera which is attached on the nose of the vehicle. The vision computer grabs target images from views of the camera, calculates its height with four feature points on the landing target. In landing mode, the helicopter follows the heading reference command first, then lands on the target by means of estimated information given by the vision computer.