Distributed Machine Learning for Terrestrial and Non-Terrestrial Internet of Things Networks

The integration of terrestrial and non-terrestrial (TNT) networks in the next-generation Internet of Things (IoT) is expected to provide a new communication paradigm for a variety of service and application offerings. In this work, we explain our vision of an intelligent and cognitive TNT-IoT platform that utilizes distributed machine learning. More specifically, we propose a novel intelligent adaptation layer (IAL) concept and examine its architecture, specifications, and operation. We investigate this new architecture's efficiency and vulnerability through a federated learning use case for joint jamming detection and waveform classification. We also identify significant research challenges and opportunities for further improvement of the TNT-IoT platform using digital twin, semantic communications, and responsible Al-empowered communications.