The landing phase is responsible for approximately36% of fatal accidents involving heavy aircraft. Traditionalinstrument landing Systems (ILS) and Global Navigation SatelliteSystems (GNSS) have demonstrated reliability and safety compliance; however, they are limited by substantial infrastructurecosts, site-specific terrain requirements, and vulnerability tosignal interference. This paper presents a comprehensive surveyof autonomous landing systems by explicitly distinguishing between certified instrument-based approaches for manned aircraftand emerging Artificial Intelligence (AI) and computer visiontechniques primarily developed for Uncrewed Aerial Vehicles(UAVs). Analyzing the transition from edge-detection techniquesto advanced deep learning models, the study highlights theircapability for infrastructure-independent navigation in GPSdenied environments. The study identifies that existing visionbased methods lack environmental robustness. Concluding, thefuture of autonomous landing relies not on replacing instrumentsystems, but on multi-modal sensor fusion, integrating GNSSpositioning with the semantic understanding of computer vision,and the advancement of Explainable AI (XAI) are necessary forfuture autonomous landings to maintain operational safety andregulatory compliance rather than replacing instrument systems.
Elashry et al. (Mon,) studied this question.