AIoT: The Conceрt, Architecture and Key Techniques

Artificial Intelligence of Things (AIoT) is an emerging research field with broad development prospects, realizing the evolution from “Internet of Everything” to “Intelligent Connection of Everything.” With the technological development such as artificial intelligence, edge computing, the Internet of Things, and mobile/embedded devices, AIoT aims to build a self-organizing, self-learning, self-adaptive, and continuous-evolving smart IoT system based on the deep fusion of these advanced technologies. AIoT is the combination of Artificial Intelligence (AI) technology and Internet of Things (IoT) infrastructure to achieve more intelligent IoT applications and provide more efficient services. Artificial intelligence models are good at analyzing and mining the potential patterns and strategies from massive amounts of data, while IoT has the ability to establish extensive connectivity for hundreds of millions of physical devices. AIoT empowers the perception, communication, computing, and application of the Internet of Things through various artificial intelligence technologies, and draws new features such as ubiquitous intelligent sensing, cloud-edge-end collaborative computing, distributed machine learning, and human-machine-things fusion. It has higher flexibility, self-organization, and self-adaptability. Specifically, AIoT enables the collection of multi-modal real-word data in real-time, and then utilizes machine learning approaches on the end devices, edge clusters or cloud servers for intelligent processing and decision making. This paper systematically introduces the direction of AIoT. Specifically, this paper firstly introduces the essential conceptual characteristics of AIoT, and then elaborates its architecture. Furthermore, we detail the research challenges and key technologies in AIoT, including ubiquitous intelligent sensing, mobile crowd sensing and computing, AIoT communication, terminal-adapted deep computing, AIoT distributed learning, cloud-edge- end collaborative computing, as well as security and privacy protection. Finally, based on the latest research developments, we present the future research directions, including collaborative soft and hard intelligence, AIoT-oriented intelligent evolution, new generation of intelligent IoT networks, continuous evolution of dynamic models, crowd intelligence with the deep fusion of Human, Machine, and IoT, and universal AIoT platforms.