Enabling Autonomous Aircraft Taxiing Navigation Through Monocular Vision

In the aviation industry, autonomous vehicles are gaining more and more importance over time, asking for a reduction of accidents by lowering the pilots' workload in the most stressful maneuvers and introducing automated systems. While autonomous flight is widely studied, being de-facto a standard in many aircraft, autonomous ground navigation is still in its early stages. With increasing air traffic, managing on-ground operations has become very challenging, especially in low-visibility conditions. Exploiting the Global Positioning System (GPS) is a possible solution. Unfortunately, this approach is prone to inaccuracies, disturbances, and jamming. For this reason, a more robust solution includes using other exteroceptive sensors, such as cameras or radars. In this paper, we propose a 2-layer control system architecture capable of performing autonomous taxiing maneuvers using a monocular camera. We propose a model-oriented approach to control the ground handling dynamics, which are used to actively exploit the information from the camera. The results, obtained using a validated multibody simulator interfaced with a graphical engine, show good tracking performance and robustness to external light conditions.