A survey on artificial intelligence‑driven data processing for IoT sensor networks

Artificial Intelligence of Things (AIoT), as the core paradigm of the deep integration of AI and IoT, has become a key force driving the global technological revolution. This paper systematically reviews the evolution and breakthrough achievements of the three core pillars of the AIoT technology system: Sensor Data Sensing, Sensor Data Networking, and Sensor Data Computing. In the data sensing layer, we focus on elaborating how AI converts physical signals from heterogeneous modalities—such as inertial, Radio Frequency (Wi-Fi/RFID/mmWave), visual, and acoustic signals—into high-level semantics, and discuss frontier methods utilizing generative models to solve the problems of data scarcity and heterogeneity. In the data networking layer, we analyze the application of AI in complex channel estimation, intelligent interference cancellation, and dynamic resource scheduling, demonstrating the evolution of networks from passive pipelines to adaptive intelligent systems. In the data computing layer, we deeply dissect the trend of computing power sinking from the cloud to the edge, covering key technologies such as on-device inference acceleration, privacy-preserving computation offloading, and federated learning. By tracing the technical innovations of key teams in various fields, this paper reveals the evolutionary logic of AIoT in solving core pain points such as acquisition accuracy, transmission reliability, and processing efficiency, providing a clear technical lineage and systematic reference for academic research and industrial applications in this field.