A framework for creating an IoT system specification with ChatGPT
Mario Binder, Vitaliy Mezhuyev
Abstract
The Internet of Things (IoT) is a core concern for the digital transformation of the industry. However, the development of IoT systems, starting from creating specifications, is a serious and time-consuming effort, which requires high levels of expertise. This paper introduces a novel approach leveraging large language models, specifically ChatGPT, to streamline the generation of IoT system specifications from initial, unstructured customer requirements. A framework, adapting the ISO IEC/IEEE 12207:2017 standard, is developed for this purpose. The utility of this framework is validated through the development and expert evaluation of a specification for a campus laboratory IoT use case. The findings demonstrate ChatGPT's effectiveness in producing specifications that are understandable, complete, and unambiguous, thus offering a substantial aid to the industry in overcoming the complexities of IoT system specification development. At the same time, the study identified specific tasks within the requirements engineering process for IoT systems that ChatGPT could not effectively perform, such as verification, obtaining agreements, and maintaining traceability of requirements between stakeholders. The findings offer valuable insights into the practical applications of AI in the domain of IoT development, particularly in automating the specification generation process. These contributions not only advance the understanding of AI's role in IoT development but also open pathways for further research on refining AI-assisted methodologies for system specification and beyond.