用自然语言指令驱动工业自动化，让机器理解人类意图。
Agentic AI for Intent-Based Industrial Automation
- 通过自然语言表达目标，自动分解为可执行任务
- 在预测性维护场景中实现自主决策与多智能体协同
- 适合希望降低技术门槛的工业界从业者
近年来，由自主大语言模型驱动的智能体系统，具备规划与工具使用能力，为工业自动化演进提供了新可能，有助于缓解工业4.0带来的复杂性。本文提出一个概念框架，将智能体AI与原本用于网络研究的意图驱动范式结合，简化人机交互（HMI），更好契合工业5.0以人为本、可持续和韧性的原则。基于意图处理，该框架允许操作员以自然语言表达高层次业务或运营目标，并将其分解为期望、条件、目标、上下文和信息等组件，由配备专用工具的子智能体执行领域特定任务。利用CMAPSS数据集与Google Agent Developer Kit（ADK）实现原型验证，展示了意图分解、智能体编排与自主决策在预测性维护场景中的可行性。结果表明，该方法有望降低技术门槛，实现可扩展的意图驱动自动化，尽管仍存在数据质量和可解释性问题。
原文摘要 · Abstract (English)
The recent development of Agentic AI systems, empowered by autonomous large language models (LLMs) agents with planning and tool-usage capabilities, enables new possibilities for the evolution of industrial automation and reduces the complexity introduced by Industry 4.0. This work proposes a conceptual framework that integrates Agentic AI with the intent-based paradigm, originally developed in network research, to simplify human-machine interaction (HMI) and better align automation systems with the human-centric, sustainable, and resilient principles of Industry 5.0. Based on the intent-based processing, the framework allows human operators to express high-level business or operational goals in natural language, which are decomposed into actionable components. These intents are broken into expectations, conditions, targets, context, and information that guide sub-agents equipped with specialized tools to execute domain-specific tasks. A proof of concept was implemented using the CMAPSS dataset and Google Agent Developer Kit (ADK), demonstrating the feasibility of intent decomposition, agent orchestration, and autonomous decision-making in predictive maintenance scenarios. The results confirm the potential of this approach to reduce technical barriers and enable scalable, intent-driven automation, despite data quality and explainability concerns.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。