数据与计算发展前沿 ›› 2026, Vol. 8 ›› Issue (2): 40-53.

CSTR: 32002.14.jfdc.CN10-1649/TP.2026.02.004

doi: 10.11871/jfdc.issn.2096-742X.2026.02.004

• 专刊：冰冻圈大数据挖掘分析关键技术及应用 • 上一篇    下一篇

基于多智能体协同的高寒山区道路结冰数据工程预警

刘景琦1,2,3(),张耀南1,2,*(),康建芳1,2,3,刘杰4,5,杨治纬4,5,张智星6,王保得7,8   

  1. 1 中国科学院西北生态环境资源研究院， 甘肃 兰州 730030
    2 国家冰川冻土沙漠科学数据中心， 甘肃 兰州 730030
    3 中国科学院大学， 北京 100049
    4 新疆交通规划勘察设计研究院有限公司， 新疆 乌鲁木齐 830006
    5 新疆高寒高海拔山区交通基础设施安全与健康重点实验室， 新疆 乌鲁木齐 830006
    6 深圳技术大学， 广东 深圳 518118
    7 中国科学院新疆生态与地理研究所， 新疆 乌鲁木齐 830011
    8 干旱区生态安全与可持续发展全国重点实验室， 新疆 乌鲁木齐 830011
  • 收稿日期:2025-10-10 出版日期:2026-04-20 发布日期:2026-04-23
  • 通讯作者: *张耀南（E-mail: yaonan@lzb.ac.cn）
  • 作者简介:刘景琦，中国科学院西北生态环境资源研究院，博士研究生，主要研究方向为数据工程防灾减灾。
    本文主要承担工作为文献梳理、论文撰写。
    LIU Jingqi is a Ph.D. student at the Northwest Institute of Eco-Environment and Resources. His main research interests include disaster prevention and reduction with data engineering.
    In this paper, he is mainly responsible for literature review and paper writing.
    E-mail: liujingqi@nieer.ac.cn|张耀南，中国科学院西北生态环境资源研究院，研究员，博士生导师，国家冰川冻土沙漠科学数据中心主任，主要研究方向为地学数据工程及数据工程防灾减灾。
    本文主要承担工作为论文修改。
    ZHANG Yaonan, Ph.D., is a professor and Ph.D. supervisor at the Northwest Institute of Eco-Environment and Resources and director of the National Cryosphere Desert Science Data Center. His main research interests include data engineering, disaster prevention and reduction with data engineering.
    In this paper, he is mainly responsible for revising the paper.
    E-mail: yaonan@lzb.ac.cn
  • 基金资助:
    国家重点研发计划(2022YFF0711704);2022年度新疆交通运输行业科技项目(2022-ZD-006);新疆交投2021年揭榜挂帅科技项目(ZKXFWCG2022060004);新疆交通设计院公司科研基金(KY2022041101);新疆维吾尔自治区科技支疆项目(2024E02030)

Multi-Agent Collaborative Data-Engineering-Based Early Warning for Road Icing in High-Altitude Cold Mountainous Regions

LIU Jingqi1,2,3(),ZHANG Yaonan1,2,*(),KANG Jianfang1,2,3,LIU Jie4,5,YANG Zhiwei4,5,ZHANG Zhixing6,WANG Baode7,8   

  1. 1 Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou, Gansu 730030, China
    2 National Cryosphere Desert Data Center, Lanzhou, Gansu 730030, China
    3 University of Chinese Academy of Sciences, Beijing 100049, China
    4 Xinjiang Transport Planning Survey and Design Institute Co. Ltd., Urumqi, Xinjiang 830006, China
    5 Xinjiang Key Laboratory for Safety and Health of Transportation Infrastructure in Alpine and High-altitude Mountainous Areas, Urumqi, Xinjiang 830006, China
    6 Sino-German College of Intelligent Manufacturing, Shenzhen Technology University, Shenzhen, Guangdong 518118, China
    7 Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Chinese Academy of Sciences, Urumqi, Xinjiang 830011, China
    8 State Key Laboratory of Ecological Security and Sustainable Development in Arid Areas, Urumqi, Xinjiang 830011, China
  • Received:2025-10-10 Online:2026-04-20 Published:2026-04-23

RichHTML

2

PDF

14

摘要：

【目的】 针对高寒山区复杂环境下传统道路结冰预警方法存在的适应性不足、协同效率低及工程化落地困难等问题，本研究旨在探索并验证一种融合多智能体协同与数据工程化的新型预警模式，提升高寒山区道路结冰预警的精准性、时效性及系统可部署性，为工程应用提供切实可行的解决方案。【方法】 引入人工智能大模型构建数据工程新模式，充分利用其强大的语义理解、多模态融合与自适应推理能力，重塑数据工程价值链。通过构建标准化、自动化的“数据-方法-分析-决策”智能工程流水线，实现从原始数据筛选、工程化分析到决策服务的全过程协同，使数据分析从一次性任务转变为可复用的工程实践。【结论】 本研究提出“智能体+数据工程”融合模式，在独库公路开展道路结冰预警模拟应用，通过持续挖掘摄像头视频数据，并融合沿线气象站点等多源异构数据，解决了道路结冰预警的精准性和时效性问题，提升了数据复用性。结果表明，该模式具备在高寒复杂环境下工程化落地的可行性，为道路结冰灾害预警提供了一条高效、可靠的新路径，有效推动了数据从“资源”向“资产”的转化。

关键词: 数据工程, 道路结冰预警, 智能体, 人工智能, 独库公路

Abstract:

[Objective] This study focuses on the limitations of traditional road icing warning methods in high-altitude mountainous environments, including poor adaptability, low coordination efficiency, and difficulties in practical implementation. It aims to explore and validate a new warning approach that integrates multi-agent collaboration with data engineering, with the goal of improving the accuracy, timeliness, and deployability of road icing warnings in such regions and providing a practical solution for engineering applications. [Methods] A new data engineering paradigm is established through the integration of large-scale artificial intelligence models, leveraging their advanced capabilities in semantic understanding, multimodal fusion, and adaptive reasoning to reconstruct the data engineering value chain. A standardized and automated “data-method-analysis-decision” intelligent pipeline is developed to enable end-to-end collaboration, encompassing data selection, engineering analysis, and decision support. This approach transforms data analysis from a one-off task into a continuously scalable and reusable engineering process. [Conclusions] This study proposes a fusion approach combining intelligent agents with data engineering and applies it to a road icing warning simulation on the G217 Duku Highway. By continuously analyzing video data from cameras and integrating multi-source heterogeneous data from along the route, this approach improves the accuracy and timeliness of road icing warnings while enhancing data reusability. The results indicate that the method is feasible for practical implementation in high-altitude, complex environments. It provides an efficient and reliable solution for road icing disaster warning and effectively promotes the transformation of data from a “resource” into a valuable “asset.”

Key words: data engineering, road icing warning, intelligent agents, AI, Duku Highway