摘要
为了改善城市路网整体通行效率,针对城市交通流实时变化以及高度非线性的特点,提出一种基于神经网络的城市关联交叉口交通流预测控制方法。以关联交叉口的总体排队长度最小为性能指标,以神经网络作为交叉口的交通流量预测模型,并通过交通流量实测值与预测值之间的误差实现对神经网络的在线训练。结合神经网络预测控制方法,给出了城市关联交叉口信号控制中信号周期、绿信比(或相位时长)以及相位差的确定思路。仿真试验结果表明:以关联交叉口总体排队长度最小为优化指标的交通流预测控制策略,可提高关联交叉口的总体通行效率。
In order to improve running efficiency of entire urban network, aimed at character of urban traffic flow which is high nonlinear and real time change, a traffic flow predictive control method of urban correlation intersection based on neural network was proposed. The principle of the method is to minimize global vehicle queue length. Neural networks were used to predict traffic volume of intersections and were trained by error between real traffic volume and predicted value. According to the given method, the computation method of cycle, split and phase offset was presented. Moreover, a simulation experiment shows that using global vehicle queue length as performance index is effective to improve running efficiency of urban network.
出处
《中国公路学报》
EI
CAS
CSCD
北大核心
2008年第5期91-95,共5页
China Journal of Highway and Transport
基金
国家高技术研究发展计划(“八六三”计划)项目(2006AA11Z211)
国家自然科学基金项目(50578064)
关键词
交通工程
关联交叉口
神经网络
交通流
预测控制方法
traffic engineering
correlation intersection
neural network
traffic flow
predictive control method
