Scholarlink

<p>Artificial intelligence (AI) for controlling mechanical systems has traditionally been used as a mathematical tool, but with the emergence of large language models (LLMs), it is evolving into agents that interact with their environment and make decisions. Based on linguistic reasoning and knowledge acquired through extensive pre-training, LLMs present a new paradigm for machine control by combining machine learning and digital twins. Notably, the approach of fine-tuning pre-trained models for specific fields and improving performance with various techniques at the inference stage has greatly expanded the capabilities of traditional AI. As demonstrated in nuclear reactor autonomy and robotic control among other cases, LLMs can serve as intelligent agents capable of interpreting abstract human instructions and translating them into executable actions within complex mechanical systems. These technological innovations are expected to go beyond mere automation, fostering a new industrial landscape where seamless collaboration between humans and machines becomes the norm. With the increasing adoption of LLM-based control systems in aerospace, shipbuilding, offshore, and heavy industries, the intelligentization of the entire industrial sector is set to occur at an accelerated rate. This paper explores how LLM-based agents, integrated with digital twins, enable autonomous and intelligent control of complex mechanical systems, focusing on nuclear reactor operations as a prime example.</p> Graphical abstract <p></p>

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

Toward intelligent mechanical systems: control and decision-making with large language models

  • 期刊论文

  • 发表日期：

    2025年6月3日

查看全文
  • Yonggyun Yu,
  • Seung Geun Kim,
  • Yoon Pyo Lee,
  • Joowon Cha

摘要

Artificial intelligence (AI) for controlling mechanical systems has traditionally been used as a mathematical tool, but with the emergence of large language models (LLMs), it is evolving into agents that interact with their environment and make decisions. Based on linguistic reasoning and knowledge acquired through extensive pre-training, LLMs present a new paradigm for machine control by combining machine learning and digital twins. Notably, the approach of fine-tuning pre-trained models for specific fields and improving performance with various techniques at the inference stage has greatly expanded the capabilities of traditional AI. As demonstrated in nuclear reactor autonomy and robotic control among other cases, LLMs can serve as intelligent agents capable of interpreting abstract human instructions and translating them into executable actions within complex mechanical systems. These technological innovations are expected to go beyond mere automation, fostering a new industrial landscape where seamless collaboration between humans and machines becomes the norm. With the increasing adoption of LLM-based control systems in aerospace, shipbuilding, offshore, and heavy industries, the intelligentization of the entire industrial sector is set to occur at an accelerated rate. This paper explores how LLM-based agents, integrated with digital twins, enable autonomous and intelligent control of complex mechanical systems, focusing on nuclear reactor operations as a prime example.

Graphical abstract