Vision-based autonomous landing of a quadrotor using a gimbaled camera

Abstract

This paper proposes a novel vision-based autonomous landing scheme for micro aerial vehicles with perturbations by using a gimbaled camera. There are no sensors available on the moving target in the task. The relative position between the drone and moving target is obtained by the camera mounted on a two-degree-of-freedom pan-tile platform. Firstly, the detection algorithm runs in real time and outputs the pixel tracking errors. Then, an adaptive vision-based controller for the pan-tilt platform guarantees that the line of sight of the camera always tracks the target. Next, the relative position tracking error is approximated according to the gimbal's rotation angles. Disturbance observer-based control is applied for flight control to attenuate the effect of disturbance and recover the hign tracking performance, where relative velocity and lumped disturbances are estimated by extend disturbance observers. The proposed flight controller guarantees that the tracking errors are ultimately bounded with tunable ultimate bounds. The convergence property is demonstrated through Lyapunov theory. The simulations and experiments illustrate the effectiveness and the superiority performance of the proposed control system.