基于MP-DDQN的智能交通信号灯控制算法  

Intelligent traffic signal control algorithm based on MP-DDQN

作　　者：王鼎盛 丁磊[1] WANG Ding-sheng;DING Lei(School of Electronic Information and Artificial Intelligence,Future Research Insititute of Integrated Circuits and Applications,Shaanxi University of Science&Technology,Xi′an 710021,China)

机构地区：[1]陕西科技大学电子信息与人工智能学院、集成电路应用与未来研究院,陕西西安710021

出　　处：《陕西科技大学学报》2025年第2期196-202,214,共8页Journal of Shaanxi University of Science & Technology

基　　金：陕西省自然科学基础研究计划项目(2020JQ-734)。

摘　　要：针对当前交通阻塞状况日益增加,传统的交通灯固定时长控制方法灵活性较差,无法根据环境自适应配时的问题,本文提出一种基于混合局部随机探索的Double DQN算法的智能交通信号灯控制方法MP-DDQN.本方法首先在Double DQN方法的基础上引入MLCA注意力机制,增强交通信号灯控制模型对于不同情景和特征的学习能力,优化交通流量的效率.其次,结合偏好引导随机探索的方法,根据当前状态的特征,有针对性地选择探索动作,减少了随机探索的开销,高效的学习到最优的交通信号灯控制策略.实验结果表明,本文提出的方法在交通流量优化方面表现较好,1000车流量中车辆的平均排队长度为2.32辆,车辆平均行驶速度为3.97m/s,相比于主流的控制方法更加高效,可以为城市交通系统的优化与改进提供有力的支持.In response to the increasing traffic congestion situation and the lack of flexibility in traditional fixed duration control methods for traffic lights,which cannot adaptively schedule according to the environment,this paper proposes an intelligent traffic signal control method MP-DDQN based on a mixed local random exploration Double DQN algorithm.This method first introduces the MLCA attention mechanism on the basis of the Double DQN method,enhancing the learning ability of the traffic signal control model for different scenarios and features,and optimizing the efficiency of traffic flow.Secondly,by combining the method of preference guided random exploration,targeted exploration actions are selected based on the characteristics of the current state,reducing the cost of random exploration and efficiently learning the optimal traffic signal control strategy.The experimental results show that the method proposed in this article performs well in traffic flow optimization,with an average queue length of 2.32 vehicles and an average driving speed of 3.97m/s in a flow of 1000 vehicles.Compared with mainstream control methods,it is more efficient and can provide strong support for the optimization and improvement of urban transportation systems.

关 键 词：交通信号灯控制 混合局部通道注意力 偏好引导 强化学习 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] U491.54[自动化与计算机技术—控制科学与工程]