知识增强大语言模型的区域交通信号控制方法

doi:10.16182/j.issn1004731x.joss.25-0921

摘要/Abstract

摘要：

自适应交通信号控制是缓解区域交通拥堵的关键，但其在突发事件实时响应与全局协同方面仍面临严峻挑战。为解决DRL方法依赖纯数据驱动，存在泛化性差、可解释性弱、缺乏突发事件处置知识指导难以满足复杂交通场景需求的问题，提出一种结合知识驱动和数据优化的控制系统。通过GraphRAG构建动态交通知识图谱，为LLM提供实时更新的历史突发事件处置经验与路网拓扑知识；采用LoRA微调技术，在避免LLM全量训练高成本的同时，精准注入交通信号控制领域专业知识；设计结构化提示词拼接区域路网状态与相关上下文，使LLM具备区域协同与事件响应能力。仿真实验表明：相比传统及强化学习基线最优方法，该方法平均通行时间分别降低9.85%、4.87%和17.4%，平均等待时间在济南与Random场景中分别降低4.17%和6.76%，吞吐量分别提升0.95%、3.71%和8.53%，且在Random场景下性能提升尤为显著。

关键词: 交通信号控制, 知识图谱, 图检索增强生成, 微调, 大语言模型, 强化学习

Abstract:

Adaptive traffic signal control (ATSC) is crucial for alleviating regional traffic congestion, yet it faces severe challenges in real-time response to unexpected events and global coordination. The DRL method relies on pure data-driven approaches, suffering from core limitations such as poor generalization, weak interpretability, and a lack of guidance from emergency disposal knowledge, which makes them difficult to meet the demands of complex traffic scenarios. A control system that integrates knowledge-driven and data-optimized approaches was proposed. The GraphRAG was used to construct a dynamic traffic knowledge graph, providing LLMs with real-time updated historical emergency disposal experience and road network topology knowledge. The LoRA fine-tuning technologywas adopted to accurately inject professional knowledge in the field of traffic signal control while avoiding the high cost of full-scale LLM training. A structured Prompt was also designed to concatenate regional road network states and relevant contexts, enabling LLMs to possess regional coordination and event response capabilities. Simulation experiments show that compared with traditional methods and the optimal reinforcement learning baseline method, this approach reduces the average travel time by 9.85%, 4.87%, and 17.4% respectively, shortens the average waiting time by 4.17% (Ji′nan dataset) and 6.76% (Random scenario) respectively, and increases the throughput by 0.95%, 3.71%, and 8.53% respectively. In addition, the performance improvement is particularly significant in the Random scenario.

Key words: traffic signal control, knowledge graph, graph retrieval-augmented generation, fine-tuning, LLMs, RL

中图分类号:

TP391

胥日升,杨林瑶,覃缘琪等 . 知识增强大语言模型的区域交通信号控制方法[J]. 系统仿真学报, 2026, 38(2): 518-531.

Xu Risheng,Yang Linyao,Qin Yuanqi,et al . Knowledge-enhanced LLM-based Method for Regional Traffic Signal Control[J]. Journal of System Simulation, 2026, 38(2): 518-531.

图/表 6

图1

图2

图3

图4

表1

图5

参考文献 41

[1]	《中国公路学报》编辑部. 中国汽车工程学术研究综述·2023[J]. 中国公路学报, 2023, 36(11): 1-192.
	Editorial Department of China Journal of Highway and Transport. Review on China's Automotive Engineering Research Progress: 2023[J]. China Journal of Highway and Transport, 2023, 36(11): 1-192.
[2]	陈虹宇, 艾红, 王晓, 等. 社会交通中的社会信号分析与感知[J]. 自动化学报, 2021, 47(6): 1256-1272.
	Chen Hongyu, Ai Hong, Wang Xiao, et al. Analysis and Perception of Social Signals in Social Transportation[J]. Acta Automatica Sinica, 2021, 47(6): 1256-1272.
[3]	杨静, 王晓, 王雨桐, 等. 平行智能与CPSS: 三十年发展的回顾与展望[J]. 自动化学报, 2023, 49(3): 614-634.
	Yang Jing, Wang Xiao, Wang Yutong, et al. Parallel Intelligence and CPSS in 30 Years: An ACP Approach[J]. Acta Automatica Sinica, 2023, 49(3): 614-634.
[4]	孙迪, 周进, 刘思佳, 等. 考虑非机动车违章影响的车辆运行效率分析[J]. 系统仿真学报, 2020, 32(11): 2166-2175.
	Sun Di, Zhou Jin, Liu Sijia, et al. Under Impact of Non-motor Vehicle Violation an Analysis on Vehicle Operation Efficiency[J]. Journal of System Simulation, 2020, 32(11): 2166-2175.
[5]	陈真, 吴卓屹, 张霖. 深度强化学习中策略表征研究简述[J]. 系统仿真学报, 2025, 37(7): 1753-1769.
	Chen Zhen, Wu Zhuoyi, Zhang Lin. Research on Policy Representation in Deep Reinforcement Learning[J]. Journal of System Simulation, 2025, 37(7): 1753-1769.
[6]	杨忠程, 叶晨, 杨振宇. 一种高效的实时交通信号灯控制算法[J]. 系统仿真学报, 2015, 27(6): 1348-1356.
	Yang Zhongcheng, Ye Chen, Yang Zhenyu. Efficient Real-time Traffic Signal Control Algorithm[J]. Journal of System Simulation, 2015, 27(6): 1348-1356.
[7]	黄秋实, 王艳阳, 邬昌良, 等. 信号交叉口混行交通协同控制方法[J]. 系统仿真学报, 2025, 37(1): 271-283.
	Huang Qiushi, Wang Yanyang, Wu Changliang, et al. Cooperative Control Method of Mixed Traffic at Signalized Intersection[J]. Journal of System Simulation, 2025, 37(1): 271-283.
[8]	覃缘琪, 季青原, 葛俊, 等. 城市交通路网动态短时推理与精准预测研究[J]. 智能科学与技术学报, 2022, 4(3): 380-395.
	Qin Yuanqi, Ji Qingyuan, Ge Jun, et al. Short-term Traffic State Reasoning and Precise Prediction in Urban Networks[J]. Chinese Journal of Intelligent Science and Technology, 2022, 4(3): 380-395.
[9]	陈苑文, 王晓, 李灵犀, 等. 基于社会媒体数据增强的交通态势感知研究及进展[J]. 智能科学与技术学报, 2022, 4(1): 1-13.
	Chen Yuanwen, Wang Xiao, Li Lingxi, et al. Traffic Situational Awareness Research and Development Enhanced by Social Media Data: the State of the Art and Prospects[J]. Chinese Journal of Intelligent Science and Technology, 2022, 4(1): 1-13.
[10]	Zhang Wei, Zhu Fenghua, Chen Yuanyuan, et al. Differential Time-variant Traffic Flow Prediction Based on Deep Learning[C]//2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC). Piscataway: IEEE, 2020: 1-6.
[11]	杜泉成, 王晓, 李灵犀, 等. 行人轨迹预测方法关键问题研究: 现状及展望[J].智能科学与技术学报, 2023, 5(2): 143-162.
	Du Quancheng, Wang Xiao, Li Lingxi, et al. Key Problems and Progress of Pedestrian Trajectory Prediction Methods: The State of the Art and Prospects[J]. Chinese Journal of Intelligent Science and Technology, 2023, 5(2): 143-162.
[12]	李炎英, 王新宇, 王晓, 等. 基于大语言模型的交通异常事件检测与辅助决策[J]. 智能科学与技术学报, 2024, 6(3): 347-355.
	Li Yanying, Wang Xinyu, Wang Xiao, et al. Traffic Anomaly Event Detection and Auxiliary Decision-making Based on Large Language Models[J]. Chinese Journal of Intelligent Science and Technology, 2024, 6(3): 347-355.
[13]	许明, 李金烨, 左东宇, 等. 基于流量预测的信号灯配时优化强化学习方法[J]. 系统仿真学报, 2025, 37(4): 1051-1062.
	Xu Ming, Li Jinye, Zuo Dongyu, et al. Signal Timing Optimization via Reinforcement Learning with Traffic Flow Prediction[J]. Journal of System Simulation, 2025, 37(4): 1051-1062.
[14]	Du Quancheng, Xu Lele, Wu Qiong, et al. SafeCrossNet: Multi-modal Fusion with Social-aware for Pedestrian Crossing Intention Prediction[J]. Information Fusion, 2026, 126, Part B: 103609.
[15]	Yoon Jinwon, Ahn Kyuree, Park Jinkyoo, et al. Transferable Traffic Signal Control: Reinforcement Learning with Graph Centric State Representation[J]. Transportation Research Part C: Emerging Technologies, 2021, 130: 103321.
[16]	Kuang Senyun, Liu Yang, Wang Xin, et al. Harnessing Multimodal Large Language Models for Traffic Knowledge Graph Generation and Decision-making[J]. Communications in Transportation Research, 2024, 4: 100146.
[17]	Yang Linyao, Chen Hongyang, Wang Xiao, et al. Open-set Entity Alignment Using Large Language Models with Retrieval Augmentation[C]//2024 IEEE International Conference on Systems, Man, and Cybernetics (SMC). Piscataway: IEEE, 2024: 4099-4104.
[18]	Lu Jingwei, Wang Xingxia, Cheng Xiang, et al. Parallel Factories for Smart Industrial Operations: from Big AI Models to Field Foundational Models and Scenarios Engineering[J]. IEEE/CAA Journal of Automatica Sinica, 2022, 9(12): 2079-2086.
[19]	Chen Yuanyuan, Yisheng Lü, Wang Xiao, et al. Traffic Flow Prediction with Parallel Data[C]//2018 21st International Conference on Intelligent Transportation Systems (ITSC). Piscataway: IEEE, 2018: 614-619.
[20]	Qin Ziye, Ji Ang, Sun Zhanbo, et al. Game Theoretic Application to Intersection Management: A Literature Review[J]. IEEE Transactions on Intelligent Vehicles, 2025, 10(4): 2589-2607.
[21]	Li Jinjue, Yu Chunhui, Shen Zilin, et al. A Survey on Urban Traffic Control Under Mixed Traffic Environment with Connected Automated Vehicles[J]. Transportation Research Part C: Emerging Technologies, 2023, 154: 104258.
[22]	Zhang Kaiwen, Cui Zhiyong, Ma Wanjing. A Survey on Reinforcement Learning-based Control for Signalized Intersections with Connected Automated Vehicles[J]. Transport Reviews, 2024, 44(6): 1187-1208.
[23]	Dai Xingyuan, Zhao Chen, Li Xiaoshuang, et al. Traffic Signal Control Using Offline Reinforcement Learning[C]//2021 China Automation Congress (CAC). Piscataway: IEEE, 2021: 8090-8095.
[24]	Han Yu, Wang Meng, Leclercq Ludovic. Leveraging Reinforcement Learning for Dynamic Traffic Control: A Survey and Challenges for Field Implementation[J]. Communications in Transportation Research, 2023, 3: 100104.
[25]	Zhang Liang, Wu Qiang, Shen Jun, et al. Expression Might Be Enough: Representing Pressure and Demand for Reinforcement Learning Based Traffic Signal Control[C]//Proceedings of the 39th International Conference on Machine Learning. Chia Laguna Resort: PMLR, 2022: 26645-26654.
[26]	Ma Zian, Xu Chengcheng, Kan Yuheng, et al. Adaptive Coordinated Traffic Control for Arterial Intersections Based on Reinforcement Learning[C]//2021 IEEE International Intelligent Transportation Systems Conference (ITSC). Piscataway: IEEE, 2021: 2562-2567.
[27]	Wang Maonan, Xiong Xi, Kan Yuheng, et al. UniTSA: A Universal Reinforcement Learning Framework for V2X Traffic Signal Control[J]. IEEE Transactions on Vehicular Technology, 2024, 73(10): 14354-14369.
[28]	Noaeen M, Naik Atharva, Goodman Liana, et al. Reinforcement Learning in Urban Network Traffic Signal Control: A Systematic Literature Review[J]. Expert Systems with Applications, 2022, 199: 116830.
[29]	Liang Xiaoyuan, Du Xunsheng, Wang Guiling, et al. A Deep Reinforcement Learning Network for Traffic Light Cycle Control[J]. IEEE Transactions on Vehicular Technology, 2019, 68(2): 1243-1253.
[30]	Ji Xuming, Bu Fan. Cooperative Trajectory Planning for Multiple CAVs via Multi-agent Reinforcement Learning[J]. Journal of Physics: Conference Series, 2025, 3097(1): 012023.
[31]	Wei Hua, Zheng Guanjie, Yao Huaxiu, et al. IntelliLight: A Reinforcement Learning Approach for Intelligent Traffic Light Control[C]//Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York: ACM, 2018: 2496-2505.
[32]	Su Haoran, Zhong Yaofeng, Chow J Y J, et al. EMVLight: A Multi-agent Reinforcement Learning Framework for an Emergency Vehicle Decentralized Routing and Traffic Signal Control System[J]. Transportation Research Part C: Emerging Technologies, 2023, 146: 103955.
[33]	Chen Chacha, Wei Hua, Xu Nan, et al. Toward A Thousand Lights: Decentralized Deep Reinforcement Learning for Large-scale Traffic Signal Control[C]//Proceedings of the Thirty-Fourth AAAI Conference on Artificial Intelligence and the Thirty-Second Conference on Innovative Applications of Artificial Intelligence and the Tenth Symposium on Educational Advances in Artificial Intelligence. Palo Alto: AAAI Press, 2020: 3414-3421.
[34]	Oroojlooy A, Nazari M, Hajinezhad D, et al. AttendLight: Universal Attention-based Reinforcement Learning Model for Traffic Signal Control[C]//Proceedings of the 34th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2020: 4079-4090.
[35]	Boukerche Azzedine, Zhong Dunhao, Sun Peng. A Novel Reinforcement Learning-based Cooperative Traffic Signal System Through Max-pressure Control[J]. IEEE Transactions on Vehicular Technology, 2022, 71(2): 1187-1198.
[36]	Wei Hua, Xu Nan, Zhang Huichu, et al. CoLight: Learning Network-level Cooperation for Traffic Signal Control[C]//Proceedings of the 28th ACM International Conference on Information and Knowledge Management. New York: ACM, 2019: 1913-1922.
[37]	Zang Xinshi, Yao Huaxiu, Zheng Guanjie, et al. MetaLight: Value-based Meta-reinforcement Learning for Traffic Signal Control[C]//Proceedings of the Thirty-Fourth AAAI Conference on Artificial Intelligence and the Thirty-Second Conference on Innovative Applications of Artificial Intelligence and the Tenth Symposium on Educational Advances in Artificial Intelligence. Palo Alto, CA, USA: AAAI Press, 2020: 1153-1160.
[38]	Hu E J, Shen Yelong, Wallis P, et al. LoRA: Low-Rank Adaptation of Large Language Models[C]//ICLR 2022 Conference. New York: ICLR, 2022: 1-13.
[39]	Edge D, Trinh H, Cheng N, et al. From Local to Global: A Graph RAG Approach to Query-focused Summarization[EB/OL]. (2025-02-19) [2025-09-01]. .
[40]	Zheng G, Xiong Y, Zang X, et al. Learning phase competition for traffic signal control[C]//Proceedings of the 28th ACM international conference on information and knowledge management. 2019: 1963-1972.
[41]	Cui Hongjun, Dong Xiushi, Su Yue, et al. Exploring Spatiotemporal Patterns of Frequently Congested Urban Road Segments Based on Multi-source Data: A Case Study of China's Super-large Cities[J]. Transportation Research Record, 2024, 2678(12): 2077-2093.

序号	相位	允许通行的车道
1	NTST	路口北方向直行和南方向直行车道
2	NLSL	路口北方向左转和南方向左转车道
3	NTNL	路口北方向直行和北方向左转车道
4	STSL	路口南方向直行和南方向左转车道
5	ETWT	路口东方向直行和西方向直行车道
6	ELWL	路口东方向左转和西方向左转车道
7	ETEL	路口东方向直行和东方向左转车道
8	WTWL	路口西方向直行和北方向左转车道