A systematic mapping and review on machine learning for non-terrestrial networks assisted Internet of Things: Enabling technologies

Non-terrestrial networks (NTNs), comprising unmanned aerial vehicles (UAVs), high-altitude platforms (HAPs), and satellites, are increasingly studied as complements to terrestrial infrastructures for extending connectivity to Internet of Things (IoT) devices and sensor networks. NTN-assisted IoT has been investigated across diverse application scenarios, including environmental monitoring, data collection, emergency response, intelligent transportation, disaster management, remote sensing, and smart city services. To address challenges such as energy constraints, dynamic network conditions, and heterogeneous communication links, a wide range of machine learning (ML) techniques have been explored, including deep learning, reinforcement learning, deep reinforcement learning, and federated learning. These methods have been applied to problems such as energy efficiency optimization, resource allocation, data collection, trajectory planning, task scheduling, and security enhancement. Given the rapid growth of this research area, with over 2000 related publications identified, developing a structured overview remains challenging. This paper addresses this need by presenting a large-scale bibliometric and thematic cluster analysis of ML-based NTN-assisted IoT research. The analysis organizes the literature into thematic clusters, examines co-occurrence patterns among enabling technologies and applications, and highlights recurring research directions and emerging topics such as edge intelligence, reconfigurable intelligent surfaces, non-orthogonal multiple access, digital twins, and explainable AI. Finally, the paper discusses current challenges and outlines potential directions for future research, with an emphasis on scalable and deployable NTN-assisted IoT systems.