Scholarlink

Smart agriculture demands software that connects sensing, control, and governance across heterogeneous assets. However, turning informal requirements into formal models for this software remains difficult, particularly deriving platform-independent models (PIM) in model-driven architecture that can be transformed into platform-specific models and code. In this paper, we automate the extraction of a PIM from requirements for smart irrigation. Our contribution is a metamodel, along with a multi-stage pipeline that constructs a PIM using large language models. In a case study, the pipeline completed model construction in 28 min, compared to two hours for the manual baseline, resulting in a 76.96% time savings and a 4.34 × productivity gain.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

LLM and Metamodeling for Model Extraction from Smart Agriculture Requirements

  • 章节

  • 发表日期：

    2026年

查看全文
  • Hamza Abdelmalek,
  • Mohammed Ait Oussouss,
  • Abdeslam Jakimi,
  • Rajae Gaamouche,
  • Rachid Saadane,
  • Abdellah Chehri

摘要

Smart agriculture demands software that connects sensing, control, and governance across heterogeneous assets. However, turning informal requirements into formal models for this software remains difficult, particularly deriving platform-independent models (PIM) in model-driven architecture that can be transformed into platform-specific models and code. In this paper, we automate the extraction of a PIM from requirements for smart irrigation. Our contribution is a metamodel, along with a multi-stage pipeline that constructs a PIM using large language models. In a case study, the pipeline completed model construction in 28 min, compared to two hours for the manual baseline, resulting in a 76.96% time savings and a 4.34 × productivity gain.