Vision-based Autonomous Landing for Micro Aerial Vehicles on Targets Moving in 3D Space

Abstract

A strategy for autonomous landing of Micro Aerial Vehicles (MAVs) on moving platforms is presented, based only on visual information from a monocular camera. The landing target is uniquely identified by previously known Augmented Reality (AR) markers, and its relative pose is estimated by visual servoing algorithms. Target trajectory in $\mathbb{R}^{3}$ is composed of planar translation and vertical oscillation, simulating a vessel that travels in foul weather. The visual feedback helps the aerial robot to track this vessel, while a trajectory planning method, based on the system's model, allows predicting its future pose. Simulated results using the ROS framework are used to verify the effectiveness of our proposed method.