基于深度学习的高速公路合流区交通速度预测

Traffic Speed Forecasting in Freeway Merging Areas Using Deep Learning Techniques

结合高速公路合流区交通流特性,建立基于长短期记忆网络(LSTM)的高速公路合流区交通速度预测模型;利用NGSIM公开数据集和仿真实验,进行速度预测模型的训练;建立了5个不同合流区段长度的速度预测模型,对实验结果进行分析,并提出高速公路合流区优化措施建议。研究结果表明,本文模型对合流区速度预测具有一定价值,可为高速公路合流区交通管理提供数据支撑与理论依据。

This paper aims to address the challenging problem of traffic speed prediction in merging areas by leveraging deep learning techniques.Initially,a traffic speed prediction model for freeway merging areas was developed,based on Long Short-Term Memory(LSTM)networks,taking into account the traffic flow characteristics specific to these areas.Subsequently,the model was trained using the NGSIM public dataset and simulation experiments.Finally,five speed prediction models for merging areas of different lengths were established,and the experimental results were analyzed,leading to recommendations for optimizing freeway merging areas.The study s findings indicate that the proposed model holds value in predicting speeds in merging areas,providing both data support and theoretical foundations for traffic management in these areas.