摘要
针对随机流量波动较大的交叉口,提出优化感应控制策略,采用正交试验方法获取最优控制参数组合.将最大排队长度作为通行需求阈值来优化感应控制逻辑,将设置的3种相位切换机制(优先排队、优先延误和固定顺序)加入感应控制参数组合中.在SUMO仿真中,模拟北京市北辰西路与科荟南路交叉口环境,采用正交试验方法筛选出不同交通流量下感应控制的最优参数组合.设计对比实验验证最优参数组合的有效性,将最优参数组合应用在深度Q学习(DQN)算法中进一步优化感应控制.结果表明,正交试验方法能够快速有效地获取最优参数组合;在低、中等交通流量下,与未使用最优参数组合的DQN算法相比,使用最优参数组合的DQN算法的收敛速度分别增加了48.14%、38.89%,平均累计车均延误分别减少了8.45%、7.09%.
Aiming at the intersection with large fluctuation of random traffic,an optimal induction control strategy was proposed,and the orthogonal test method was used to obtain the optimal combination of control parameters.The maximum queuing length was used as the traffic demand threshold to optimize the induction control logic,and the three phase switching mechanisms(priority queuing,priority delay and fixed order)were added to the induction control parameter combination.In the SUMO simulation,the intersection environment of Beichen West Road and Kehui South Road in Beijing was simulated,and the optimal parameter combination of induction control under each traffic flow was selected by using the orthogonal test method.A comparative experiment was designed to verify the effectiveness of the optimal parameter combination,and the optimal parameter combination was applied to the deep Q-network(DQN)algorithm to further optimize the induction control.Results show that the optimal parameter combination can be obtained quickly and effectively by using the orthogonal test method.Under the low and the medium traffic flow,compared with the DQN algorithm without optimal parameter combination,the convergence speed of the DQN algorithm using the optimal parameter combination increase by 48.14%and 38.89%respectively,and the average cumulative vehicle delay decrease by 8.45%and 7.09%respectively.
作者
王志建
龙顺忠
李颖宏
WANG Zhi-jian;LONG Shun-zhong;LI Ying-hong(School of Electrical and Control Engineering,North China University of Technology,Beijing 100144,China)
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2023年第6期1128-1136,共9页
Journal of Zhejiang University：Engineering Science
基金
国家自然科学基金资助项目(72071003)。
关键词
信号交叉口
感应控制
影响参数
正交试验
深度Q学习(DQN)算法
signalized intersection
induction control
influencing parameter
orthogonal test
deep Q-network(DQN)algorithm
