Machine Learning in Non-Terrestrial Networks for the Internet of Things

This book explains how machine learning supports non-terrestrial-network-assisted Internet of Things systems across eight application domains. Its approach combines a six-layer model, bibliometric mapping, and cross-domain examination of platform roles, learning and optimization methods, system objectives and decision variables, evaluation settings, limitations, and reported outcomes.

Its scope covers data collection; forecasting and prediction; disaster management and emergency response; smart cities; precision agriculture; remote sensing and environmental monitoring; real-time and time-sensitive operation; and Age-of-Information-aware systems. It examines how unmanned aerial vehicles, high-altitude platforms, satellites, terrestrial infrastructure, and edge and cloud resources support sensing, access, relaying, backhaul, computation, model training, inference, scheduling, resource allocation, trajectory control, and application decision-making.

This book presents a six-layer model covering sensing and actuation, non-terrestrial platforms, connectivity, computation, learning and control, and applications. Representative studies are examined according to platform role, learning method, system objective, controlled variables, constraints, metrics, baselines, and evaluation settings. Bibliometric mapping identifies recurring keywords, research themes, platforms, and learning methods within each application domain.

The reviewed studies reflect differences in platform type, learning method, application function, and evaluation conditions. This book draws lessons from analytical evaluations, simulations, public-dataset studies, prototypes, testbeds, field experiments, and operational deployments. This highlights differences in research problems, platforms, datasets, channel and mobility models, objectives, metrics, and baselines across a wide range of use cases.

Recurring technical themes, methods, and research directions span trajectory planning, radio and computing resource allocation, mobile edge computing, federated learning, deep reinforcement learning, energy management and consumption minimization, satellite connectivity and intermittent operation, security, privacy, reliability, latency, and information-freshness optimization. Limitations, deployment challenges, and future research directions are identified within each application domain.

Examines how Machine Learning transforms Non-terrestrial networkassisted IoT applications Uses Vosviewer-based data to provide a lateral mapping of all major aspects surrounding each application Presents various applications, one for each major chapter, with a comprehensive lateral taxonomy useful as a reference

Inhalt

Introduction and Methodological Framework.- Machine Learning for NTN Assisted IoT Data Collection.- Machine Learning for Forecasting and Prediction in NTN Assisted IoT.- Machine Learning for NTN Assisted Disaster Management and Emergency Response.- Machine Learning for NTN Assisted Smart Cities and Urban IoT.- Machine Learning for NTN Assisted Precision Agriculture and Smart Farming.- Machine Learning for NTN Assisted Remote Sensing and Environmental Monitoring.- Machine Learning for Real Time and Time-Sensitive NTN Assisted IoT.- Machine Learning for AoI Aware NTN Assisted IoT.