Scholarlink

<p>In complex urban traffic networks, ensuring rapid emergency vehicle (EV) passage while minimizing disruptions to ordinary traffic remains a major challenge. This study presents an intelligent dual-level system that integrates the Cell Transmission Model (CTM) for macroscopic traffic state estimation and Deep Reinforcement Learning (DRL) for intersection-level adaptive signal coordination. The upper level dynamically plans EV routes based on CTM-predicted flow and density distributions under partially connected vehicle (CV) environments, minimizing EV travel time while containing network-wide delay. The lower level employs DRL agents at signalized intersections to allocate green-time ratios among competing phases through adaptive signal splitting, responding autonomously to evolving traffic conditions and predicted EV arrivals. The two levels interact through an iterative rolling-horizon framework, where route optimization and signal adjustment continuously refine each other in real time. Extensive simulation experiments conducted on realistic urban networks demonstrate that the proposed framework reduces EV travel time by up to 28.12% and average normal-traffic delay by 30.97% compared with existing MPC-based and rule-based preemption strategies. The results verify that combining model-based CTM estimation with data-driven DRL control offers a scalable and adaptive solution for emergency response operations in mixed traffic environments.</p>

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

An Intelligent Dual-level System for Emergency Vehicle Routing and Adaptive Signal Coordination

  • 期刊论文

  • 发表日期：

    2026年2月12日

查看全文
  • Yulu Dai,
  • Liang Hu

摘要

In complex urban traffic networks, ensuring rapid emergency vehicle (EV) passage while minimizing disruptions to ordinary traffic remains a major challenge. This study presents an intelligent dual-level system that integrates the Cell Transmission Model (CTM) for macroscopic traffic state estimation and Deep Reinforcement Learning (DRL) for intersection-level adaptive signal coordination. The upper level dynamically plans EV routes based on CTM-predicted flow and density distributions under partially connected vehicle (CV) environments, minimizing EV travel time while containing network-wide delay. The lower level employs DRL agents at signalized intersections to allocate green-time ratios among competing phases through adaptive signal splitting, responding autonomously to evolving traffic conditions and predicted EV arrivals. The two levels interact through an iterative rolling-horizon framework, where route optimization and signal adjustment continuously refine each other in real time. Extensive simulation experiments conducted on realistic urban networks demonstrate that the proposed framework reduces EV travel time by up to 28.12% and average normal-traffic delay by 30.97% compared with existing MPC-based and rule-based preemption strategies. The results verify that combining model-based CTM estimation with data-driven DRL control offers a scalable and adaptive solution for emergency response operations in mixed traffic environments.