LLM-Driven Automated IoT Application Development Technology

With the continuous advancement of internet of things (IoT) technology, systems are rapidly evolving toward physical intelligence and AI-native cyber-physical systems (CPS). In this process, high fragmentation in both software and hardware has become a major obstacle to achieving system-level intelligence. IoT devices exhibit significant differences in hardware architecture, operating systems, and communication protocols, often imposing substantial complexity on developers during cross-platform adaptation. At the same time, strict constraints on computing power, storage, and power consumption further exacerbate the difficulty of system development. To address these challenges, LLM-driven automated IoT application development technologies have emerged in recent years. These technologies deeply integrate semantic understanding with physical mechanisms, enabling the automatic generation of high-quality program code based on developer intent. This approach not only significantly lowers the barrier to development but also effectively improves development efficiency. This article systematically elaborates on the automated implementation path for IoT application development from three key dimensions. In terms of device-side code generation, the research focus lies in the accurate integration of underlying library functions and the optimization of execution efficiency in resource-constrained environments. For constructing business logic on the cloud and edge sides, this article emphasizes the analysis of coordination mechanisms among multiple nodes and intelligent scheduling strategies for rule engines. Furthermore, at the level of intelligent data processing algorithms and model generation, as core components for improving system perception and decision-making capabilities, systematic evaluation of their generation quality and physical consistency is required. This article aims to present the latest research progress in the field of automated IoT application generation using LLMs, providing key technical support for advancing IoT toward an AI-native paradigm of cyber-physical systems. Through the above pathways, it is expected that next-generation physical intelligence systems with higher efficiency, evolvability, and closed-loop feedback capabilities can be constructed.