Modern transportation systems must include dynamic traffic management, and sophisticated algorithms like Hypergraph Neural Networks (HGNNs) and Multi-Agent Reinforcement Learning (MARL) are essential for maximizing traffic flow and minimizing congestion. Multiple agents, including traffic lights, vehicles, and sensors, collaborate to optimize traffic conditions throughout a network in MARL, an extension of Reinforcement Learning. To dynamically adjust to real-time circumstances, such as road closures or accidents, each agent learns independently while interacting with other agents. The system can effectively manage complex, interconnected urban traffic networks due to its decentralized design, which also enhances scalability. MARL significantly improves congestion mitigation by enabling real-time traffic signal adjustments and adaptive rerouting strategies, ultimately reducing delays and optimizing road network efficiency. By representing road networks as hypergraphs, in which multiple roads (nodes) connect via hyperedges to capture higher-order relationships, Hypergraph Neural Networks (HGNNs) extend standard Graph Neural Networks. This feature allows HGNNs to effectively analyse and predict complex traffic patterns and the propagation of congestion. When combined with temporal modelling techniques such as Temporal Convolutional Networks (TCNs), HGNNs can predict future traffic conditions, enabling proactive management strategies such as rerouting vehicles before congestion occurs. These algorithms offer flexible and scalable traffic control systems by leveraging historical trends, real-time data from GPS, and Internet of Things (IoT) sensors. The integration of MARL and HGNNs provides a robust framework for dynamic traffic optimization, fostering sustainable urban mobility, reducing emissions, and enhancing overall traffic flow. These advancements lay the foundation for future data-driven, intelligent transportation systems.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Real-Time Traffic Optimization: A Graph-Based Reinforcement Learning Framework


    Contributors:
    Sumathi, N. (author) / Navamani, G. (author)


    Publication date :

    2025-04-23


    Size :

    571508 byte





    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English



    Multi-machine real-time three-dimensional conflict resolution method based on graph reinforcement learning

    LI GUIYI / LING-LONG SIYU / OUYANG YUXIANG et al. | European Patent Office | 2025

    Free access

    Self-adaptive traffic signal control method based on graph deep reinforcement learning

    ZHAO ZHONGNAN / WANG KUN / SONG XIN | European Patent Office | 2022

    Free access

    Traffic Signal Control Based on Reinforcement Learning with Graph Convolutional Neural Nets

    Nishi, Tomoki / Otaki, Keisuke / Hayakawa, Keiichiro et al. | IEEE | 2018