Architectures And Applications Of AI-Enabled Internet Of Things (AIoT) Systems

The convergence of artificial intelligence (AI) and the Internet of Things (IoT) has given rise to AI-Enabled Internet of Things (AIoT) systems that tightly integrate large-scale sensing infrastructures with intelligent, data-driven decision-making capabilities across heterogeneous environments. In contrast to conventional IoT architectures that depend predominantly on centralized cloud analytics, AIoT systems distribute learning, inference, and control functions across device, edge, fog, and cloud layers, enabling low-latency responses, reduced network bandwidth consumption, improved energy efficiency, and enhanced data privacy. This distributed intelligence paradigm builds upon foundational advances in wireless sensor networks, fog and edge computing, and machine learning for resource-constrained environments, allowing computation to be placed closer to data sources while still leveraging cloud-scale training and orchestration. This article presents a comprehensive review of AIoT system architectures and applications, examining architectural patterns, data flows, and deployment trade-offs that govern performance, scalability, and resilience. Representative application domains including smart cities, healthcare systems, industrial automation, and intelligent transportation are discussed to illustrate how AIoT enables real-time perception, predictive analytics, and autonomous control. Finally, the article outlines open challenges and future research directions, such as scalable model management, end-to-end security and privacy, interoperability, and adaptive intelligence at the edge, which remain critical to the widespread adoption of AIoT systems.