Scholarlink

The promotion and popularization of intelligent networked vehicles has driven the change of intersection control strategy from “passive response” to “active cooperation”, and the existing research has the problems of high complexity of cooperative optimization modeling and difficult to update and optimize the strategy. How to realize more intelligent and systematic control of intersections in the complex and changing traffic environment has become an urgent problem to be solved. In this paper, the intersection traffic cooperative control problem is defined as a multi-intelligence reinforcement learning task, and the ecological guidance cooperative control architecture of networked vehicles based on hierarchical reinforcement learning is constructed, which characterizes the networked vehicles and intersection signals and their intrinsic associations through the design of heterogeneous intelligences, and the hierarchical reinforcement learning algorithm is designed for solving the ecological objectives of the intersection to solve the integrated and optimal cooperative control scheme. Taking a four-phase intersection as an example, the effectiveness of this paper’s method is verified through experimental analysis.

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

Hierarchical Reinforcement Learning-Based Cooperative Control Approach for Ecological Guidance of Networked Vehicles at Intersections

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Xiaotian Lu,
  • Yimai Zhang,
  • Haitao Li,
  • Xue Ao,
  • Pengju Liu

摘要

The promotion and popularization of intelligent networked vehicles has driven the change of intersection control strategy from “passive response” to “active cooperation”, and the existing research has the problems of high complexity of cooperative optimization modeling and difficult to update and optimize the strategy. How to realize more intelligent and systematic control of intersections in the complex and changing traffic environment has become an urgent problem to be solved. In this paper, the intersection traffic cooperative control problem is defined as a multi-intelligence reinforcement learning task, and the ecological guidance cooperative control architecture of networked vehicles based on hierarchical reinforcement learning is constructed, which characterizes the networked vehicles and intersection signals and their intrinsic associations through the design of heterogeneous intelligences, and the hierarchical reinforcement learning algorithm is designed for solving the ecological objectives of the intersection to solve the integrated and optimal cooperative control scheme. Taking a four-phase intersection as an example, the effectiveness of this paper’s method is verified through experimental analysis.