基于改进深度强化学习的交通信号灯控制  

Traffic signal control based on improved deep reinforcement learning

作　　者：韦敏[1] 蔡常健 WEI Min;CAI Chang-jian+(School of Electronic Engineering,Xi’an Shiyou University,Xi’an 710000,China)

机构地区：[1]西安石油大学电子工程学院,陕西西安710000

出　　处：《计算机工程与设计》2025年第3期927-933,共7页Computer Engineering and Design

基　　金：陕西省重点研发计划基金项目(2018GY-181、2023-YBGY-219);西安石油大学研究生创新与实践能力培养计划基金项目(YCS23213116)。

摘　　要：为解决复杂交通情境下传统交通信号灯控制效果有限的问题,提出一种改进的深度强化学习交通信号灯控制方法。将对决网络和双Q学习结合,改进深度强化学习模型结构,缓解算法的高估;设计能提取更丰富交通信息的多特征状态空间,考虑车辆等待时间和车道最大队列长度的多任务奖励函数,提高城市交叉口的通行效率。实验结果表明,所提方法能够获得更高奖励,在训练场景对比基线方法平均等待时间和平均队列长度均明显降低,平均速度明显提高,测试结果同样验证所提方法更能提高道路通行效率。An improved deep reinforcement learning method was proposed to address the limitations of traditional traffic signal control in complex traffic scenarios.The dueling network and double Q-learning were combined,enhancing the structure of the deep reinforcement learning model and mitigating algorithm overestimation.A multi-feature state space was designed to extract richer traffic information,incorporating a multi-task reward function that considered vehicle waiting time and maximum queue length per lane,to improve the efficiency of urban intersection traffic flow.Experimental results demonstrate that the proposed method achieves higher rewards,with significantly reduced average waiting times and queue lengths compared to baseline methods in training scenarios.The average speed noticeably increases,and testing results further confirm the enhanced efficiency of road traffic flow achieved using the proposed method.

关 键 词：深度强化学习 信号灯控制 对决网络 状态空间 奖励函数 城市交叉口 交通工程 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]