基于组合深度学习的快速路车道级 速度预测研究

随着物联网、云计算和大数据在智能交通领域的普及应用，传统的以道路断面为研究对象的预测方法已经无法满足智能网联技术发展的需求.本文以车道断面为研究对象，提出一种基于组合深度学习(Combined Deep Learning，CDL)的城市快速路车道级速度预测模型.该模型利用基于信息熵的灰色关联分析提取空间特征变量，采用长短期记忆神经网络提取空间特征变量的时间特征，并利用门限递归单元神经网络得到预测结果.通过北京市东二环路车道断面实测微波数据验证发现，提取车道交通流的时空特征，CDL模型能够很好地拟合不同车道不同时段的速度变化趋势，可有效地实现车道速度的单步及多步预测，且该模型的预测精度和稳定性均优于传统预测模型.

Lane-level Traffic Speed Prediction for Expressways Based on A Combined Deep Learning Model

With the increasing application of internet of things, cloud computing and big data in the field of intelligent transportation system, traditional traffic prediction methods which take the road sections as research object cannot satisfy the development of intelligent connected technique. To forecast the traffic state of lanes, a novel combined deep learning (CDL) model is proposed to predict the travel speed of the lanes of expressways. First, the CDL model introduces an entropy- based grey relation analysis to extract the variables of spatial characteristics. Then, the CDL model uses long short- term memory neural network to capture the temporal characteristics of the extracted spatial variables. Finally, the gated recurrent unit neural network is utilized to predict the travel speed of target lane section in the next time intervals. Validated by the ground-truth microwave data of lane sections on the 2nd ring road of Beijing, the proposed model can well capture the trend of speed change during different time periods of the different lanes and realize the single-step and multi-step prediction of lane speed effectively. The prediction results illustrate that the CDL model outperforms many traditional methods in terms of accuracy and stability.