智能网联环境下基于混合深度学习的交通流预测模型

为适应未来智能网联环境下精细化交通流预测需求，提出一种基于混合深度学习 (Hybrid Deep Learning, HDL)的车道级交通流速度预测模型. 模型以智能网联系统强大的数据采集和计算能力为基础，采用集成经验模态分解算法将原始速度序列分解为多个固有模态函数分量和残差分量，并将所得分量重构为模型输入；利用双向长短期记忆神经网络和注意力机制，构建深度学习模型框架；为检验模型预测精度和可靠性，选择北京市二环路多个连续车道断面速度数据进行算法验证. 结果表明，HDL模型在不同车道均有理想的预测结果，单步和多步预测精度均显著优于对比模型.

Traffic Flow Prediction Based on Hybrid Deep Learning Under Connected and Automated Vehicle Environment

To achieve refined traffic flow prediction under connected and automated vehicle highway (CAVH) environment, this study proposes a lane- level traffic flow prediction model based on the hybrid deep learning (HDL). The proposed method takes the advantages of powerful data collection and calculation capability of the CAVH system. The HDL model divided the raw traffic speed series into several intrinsic mode function components and one residual component, and used the components as the input of the model. The bidirectional long short-term memory neural network and attention mechanism were used to establish the framework of the deep learning model. The lane-level speeds of the 2nd Ring road in Beijing, China were utilized to examine the accuracy and reliability of the proposed model. The results illustrate that the HDL model has ideal prediction performance at different types of lanes. Meanwhile, the prediction accuracy of the HDL model is significantly higher than that of previous models in terms of single-step-ahead prediction and multi-step-ahead prediction.