欢迎访问《汽车安全与节能学报》，
汽车安全与节能学报

汽车安全与节能学报 ›› 2026, Vol. 17 ›› Issue (1): 1-17.DOI: 10.3969/j.issn.1674-8484.2026.01.001

• 综述与展望 •    下一篇

从解耦到协同：智能网联汽车数据与计算调度范式演进

袁宏1,2(), 黄开胜1,2, 田光宇1,*()   

  1. 1.清华大学 车辆与运载学院，北京 100084，中国
    2.浙江清华长三角研究院，绿色出行研究中心，嘉兴 314006，中国
  • 收稿日期:2025-11-11 修回日期:2026-01-14 出版日期:2026-02-28 发布日期:2026-03-19
  • 通讯作者: 田光宇，教授。E-mail：tiangy@tsinghua.edu.cn。
    清华大学教授，博士生导师。担任清华大学车辆与运载学院学术委员会主任、中国汽车工程学理事、中国汽车工程学会电动汽车分会主任委员等职务。重点研究领域包括新能源智能汽车集成与控制、电机变速器集成驱动系统控制、分布式驱动车辆动力学控制以及电池系统应用技术。
  • 作者简介:袁宏（1974—），女（汉），北京，高级工程师。E-mail：yuanhong611@mail.tsinghua.edu.cn。
  • 基金资助:
    国家重点研发计划资助(2021YFB2501000)

From decoupling to synergy: A paradigm shift in data and computation co-scheduling for intelligent connected vehicles

YUAN Hong1,2(), HUANG Kaisheng1,2, TIAN Guangyu1,*()   

  1. 1. School of Vehicle and Mobility, Tsinghua University, Beijing 100084, China
    2. Green Travel Research Center, Yangtze Delta Region Institute of Tsinghua University, Jiaxing 314006, China
  • Received:2025-11-11 Revised:2026-01-14 Online:2026-02-28 Published:2026-03-19
  • Contact: Prof. Tian GuangyuHe is a tenured professor and doctoral supervisor at Tsinghua University, the Chair of the Academic Committee of the School of Vehicle and Mobility, Tsinghua University, the Council Member of the Chinese Society of Automotive Engineers, and the Chair of the Electric Vehicle Committee of the Chinese Society of Automotive Engineers. His research interests include new energy intelligent vehicle integration and control, automatic transmission, motor control, and battery system application technologies.

RichHTML

14

PDF

81

摘要：

智能汽车信息物理系统（IVCPS）是突破单车智能局限的关键，但其性能受限于海量数据需求与动态稀缺的通信计算资源间的冲突，而该冲突源于数据流调度与计算任务调度间的强耦合关系。现有研究多采用独立优化二者的解耦方法，忽视了由此引发的系统性性能瓶颈，且缺乏一个综合性的数据-计算协同设计框架。因此，该文系统性地梳理了IVCPS中从资源驱动的独立优化到任务驱动的一体化协同这一调度范式的转变；深入剖析了从显式协调到隐式融合的协同机制演进路径，尤其是在应用多智能体系统强化学习解决分布式资源冲突及保障人工智能（AI）决策可信度方面，识别出未来的关键研究方向。该研究旨在为异构多体交互协同中数据-计算协同调度这一核心问题建立清晰的理论框架，为下一代高级别自动驾驶与智能交通系统的架构设计提供关键的理论和技术支撑。

关键词: 智能网联汽车, 数据-计算协同调度, 任务驱动, 信息物理系统（CPS）, 多智能体系统（MAS）

Abstract:

Intelligent vehicle cyber-physical systems (IVCPS) are pivotal for transcending the limitations of single-vehicle intelligence, yet their performance is constrained by the conflict between massive data demands and dynamic, scarce communication and computation resources. This conflict stems from the strong coupling between data flow scheduling and computation task scheduling. Prevailing research often adopts a decoupled approach by optimizing these two aspects independently, overlooking the resultant systemic performance bottlenecks and lacking a comprehensive framework for Data-Computation Co-Scheduling. Therefore, this paper systematically reviews the paradigm shift in IVCPS scheduling from resource-driven independent optimization to task-driven integrated co-design. It dissects the evolution of coordination mechanisms, from explicit coordination to implicit fusion, and identifies key future research directions, particularly in applying multi-agent reinforcement learning to resolve distributed resource conflicts and ensuring the trustworthiness of artificial intelligence (AI) decisions. This study aims to establish a clear theoretical framework for the core issue of data-computation co-scheduling, providing crucial theoretical and technical support for the architectural design of next-generation intelligent transportation systems and advanced autonomous driving.

Key words: Intelligent connected vehicles, data-computation co-scheduling, task-driven, cyber-physical systems (CPS), multi-agent systems (MAS)

中图分类号: