基于交通路网状态感知的交通信号配时优化技术  

Traffic Signal Timing Optimization Based on Traffic Network State Perception

作　　者：胡小光 胡晓杰 HU Xiaoguang;HU Xiaojie(Shenyang Ligong University,Shenyang Liaoning 110159)

机构地区：[1]沈阳理工大学,辽宁沈阳110159

出　　处：《河南科技》2021年第16期18-21,共4页Henan Science and Technology

摘　　要：本文提出多通道的3D卷积预测网络结合Double-DQN的配时控制策略,并将其与Deep-Q-learning、Q-learning进行对比分析,进行仿真实验。结果表明,多通道的3D卷积预测网络结合Double-DQN的控制策略明显优于Q-learning和Deep-Q-learning。In this paper,a multi-channel 3D convolution prediction network combined with Double-DQN timing control strategy was proposed,which was compared with Deep-Q-Learning and Q-Learning.The results show that the control strategy of multi-channel 3D convolution prediction network combined with Double-DQN timing control strategy is obviously better than Q-learning and Deep-Q-learning.

关 键 词：智能交通信号配时 3D卷积 深度强化学习 多通道矩阵 Double-DQN 

分 类 号：U491.54[交通运输工程—交通运输规划与管理]