The invention discloses an expressway traffic state estimation method based on a Transform model, and the method comprises the steps: collecting traffic data of an expressway, carrying out the preprocessing of the traffic data, and converting the historical traffic data into a traffic state feature vector; a traffic state estimation model based on Transform is constructed, a multi-head self-attention mechanism is utilized to capture space-time characteristics of traffic data, a deep learning model is constructed to learn a hidden mode in the traffic data, and traffic state feature vectors are utilized to train a traffic state estimator; when state estimation is carried out, the method inputs real-time traffic data into a Transform model to obtain feature vectors of the real-time traffic data, and current traffic state parameters of the expressway are obtained through the trained traffic state estimator. According to the method, the accuracy of traffic state estimation is improved, an efficient and reliable tool is provided for expressway managers, the expressway managers are helped to carry out traffic management and scheduling better, and therefore the traffic mobility and the overall performance of a road network are optimized.
本发明公开了一种基于Transformer模型的高速公路交通状态估计方法,包括:采集高速公路的交通数据并进行预处理,将历史交通数据转化为交通状态特征向量;构建基于Transformer的交通状态估计模型,利用多头自注意力机制捕获交通数据的时空特点,构建深度学习模型来学习交通数据中的隐藏模式,利用交通状态特征向量训练一个交通状态估计器;在进行状态估计时,该方法会将实时交通数据输入到Transformer模型中,获取其特征向量,并通过前述训练的交通状态估计器得到高速公路的当前交通状态参数。本发明方法不仅提高了交通状态估计的准确性,还为高速公路管理者提供了一种高效、可靠的工具,帮助他们更好地进行交通管理和调度,从而优化路网的交通流动性和整体性能。
Transform-based highway traffic state estimation method
基于Transformer的高速公路交通状态估计方法
2024-02-20
Patent
Electronic Resource
Chinese
ETC data-based highway traffic state dynamic estimation method
European Patent Office | 2024
|Highway traffic flow prediction method based on Transform and graph attention network
European Patent Office | 2023
|Highway traffic guidance method and highway traffic guidance system
European Patent Office | 2015
|Highway traffic state robust estimation method based on network-connected vehicle trajectory data
European Patent Office | 2024
|