Scholarlink

This study investigates the potential of large language models (LLMs) and intelligent agent technology in the field of automobile maintenance. With the development of artificial intelligence, the automotive maintenance process is gradually shifting from traditional manual operation to the use of AI agents to complete tasks. Especially in vehicle fault diagnosis, troubleshooting, and customer service, the use of LLMs and intelligent agent technology can significantly improve work efficiency. In this work, we developed an automotive fault-diagnosis agent system that incorporates large language models (LLMs) for analyzing user-supplied multimodal information queries and performs expert-like reasoning and questioning in the troubleshooting process, in order to make precise diagnostic and maintenance recommendations for vehicle repair and maintenance. In addition, in this study, we also evaluated the performance of various LLMs such as GPT, Gemini, Llama, and DeepSeek based models on system implementation. Based on several criteria, we also compare the applicability and performance of the LLM models for this application. The experimental results show that the developed system is capable of performing automotive fault diagnosis and providing appropriate maintenance recommendations.

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

A LLM-Driven Agent System for Automotive Fault Diagnosis with Integrated Reasoning and Expertise Sharing

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Tzu-Ting Weng,
  • Jun-Teng Sun,
  • Zhen-Xin Fu,
  • Chung-Hong Lee

摘要

This study investigates the potential of large language models (LLMs) and intelligent agent technology in the field of automobile maintenance. With the development of artificial intelligence, the automotive maintenance process is gradually shifting from traditional manual operation to the use of AI agents to complete tasks. Especially in vehicle fault diagnosis, troubleshooting, and customer service, the use of LLMs and intelligent agent technology can significantly improve work efficiency. In this work, we developed an automotive fault-diagnosis agent system that incorporates large language models (LLMs) for analyzing user-supplied multimodal information queries and performs expert-like reasoning and questioning in the troubleshooting process, in order to make precise diagnostic and maintenance recommendations for vehicle repair and maintenance. In addition, in this study, we also evaluated the performance of various LLMs such as GPT, Gemini, Llama, and DeepSeek based models on system implementation. Based on several criteria, we also compare the applicability and performance of the LLM models for this application. The experimental results show that the developed system is capable of performing automotive fault diagnosis and providing appropriate maintenance recommendations.