Hybrid GNSS-Visual Navigation for Resilient Positioning in Low Earth Orbit

Navigation refers to the process of monitoring the motion of land, marine, aeronautical and space vehicles. With the advent of Global Navigation Satellite System (GNSS), this process has achieved remarkable positioning accuracy and is now widely accessible through embedded devices, equipped with a GNSS receiver acquiring Radio Frequency (RF) signals transmitted by multiple GNSS constellations. However, the more these signals become integrated into our daily lives, the more attractive they become as targets for malicious actors. In recent years, the occurrence of intentional RF disruptions has significantly grown taking the form of jamming and spoofing attacks, which aim to degrade or deny the reception of GNSS signals. This issue is no longer confined to terrestrial applications as satellites operating in Low Earth Orbit (LEO) are increasingly exposed to these threats, potentially leading to significant errors in on-board position estimates compromising mission performance. To mitigate such risks, sequential filtering techniques are commonly employed, relying on approximate models of system dynamics and sensor measurements. In this context, the integration of complementary sensors represents a promising solution to enhance navigation robustness, such as on-board cameras for Earth Observation (EO). For this purpose, this thesis presents a comprehensive review of visual-aided navigation techniques for aerial platforms, along with recent advancements in Artificial Intelligence (AI) techniques applied to EO. Furthermore, a simulation framework has been developed to investigate the fusion of GNSS and visual measurements using an Extended Kalman Filter (EKF). The proposed technique is evaluated to assess the contribution of visual sensing to navigation accuracy and robustness.