Scholarlink

<p>Traffic jams and congestion have emerged as pressing problems for urban management due to the rapid proliferation of vehicles. Over the past few years, significant progress has been made in applying deep reinforcement learning to traffic signal control. However, current centralized control methods tend to utilize a single strategy and still exhibit some shortcomings when dealing with multi-intersection environments, such as unstable collaborative control effects and difficulties in adapting to local traffic changes. In response to this issue, an improved Independent Advantage Actor-Critic(IA2C) traffic signal control method called Dynamic Cooperative Advantage Actor-Critic(DCA2C) has been proposed in this paper. First, DCA2C uses the redesigned phase pressure information as a state input while incorporating the state and strategy information from neighboring intersections to speed up model convergence. In addition, a pressure-based dynamic cooperation factor adjustment mechanism is designed to adaptively modify the cooperation weights between intersections, enhancing their cooperative control effect. Finally, experiments are conducted utilizing both a synthetic large-scale traffic grid and a real-world traffic scenario in Monaco. The experimental results demonstrate that DCA2C performs better than the current state-of-the-art distributed Multi-Agent Reinforcement Learning methods in both scenarios, effectively reducing traffic congestion and improving the overall efficiency of the traffic network.</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

DCA2C: Dynamic cooperative A2C-based traffic signal control in multi-intersection environments

  • 期刊论文

  • 发表日期：

    2026年2月28日

查看全文
  • Qinghan Huang,
  • Lei Nie,
  • Dandan Qi,
  • Haizhou Bao

摘要

Traffic jams and congestion have emerged as pressing problems for urban management due to the rapid proliferation of vehicles. Over the past few years, significant progress has been made in applying deep reinforcement learning to traffic signal control. However, current centralized control methods tend to utilize a single strategy and still exhibit some shortcomings when dealing with multi-intersection environments, such as unstable collaborative control effects and difficulties in adapting to local traffic changes. In response to this issue, an improved Independent Advantage Actor-Critic(IA2C) traffic signal control method called Dynamic Cooperative Advantage Actor-Critic(DCA2C) has been proposed in this paper. First, DCA2C uses the redesigned phase pressure information as a state input while incorporating the state and strategy information from neighboring intersections to speed up model convergence. In addition, a pressure-based dynamic cooperation factor adjustment mechanism is designed to adaptively modify the cooperation weights between intersections, enhancing their cooperative control effect. Finally, experiments are conducted utilizing both a synthetic large-scale traffic grid and a real-world traffic scenario in Monaco. The experimental results demonstrate that DCA2C performs better than the current state-of-the-art distributed Multi-Agent Reinforcement Learning methods in both scenarios, effectively reducing traffic congestion and improving the overall efficiency of the traffic network.