The invention discloses an urban road traffic flow perception prediction method and system, the system comprises a data layer, a conversion layer and a representation layer, the data layer is located at the bottom layer of a traffic flow perception prediction system framework, the conversion layer is an intermediate layer for communicating the data layer and the representation layer, and the representation layer is located at the top layer of the system framework. The characterization layer model carries out self-supervised pre-training by using the traffic signal space-time sequence data, and extracts the space-time characteristics of the traffic road network monitoring signals. The multi-layer self-attention traffic prediction model comprises: 1) a road network space encoder which captures space information of a single-frame traffic network; 2) a traffic sequence time encoder which captures time correlation of a multi-frame traffic road network; and 3) a feature aggregator for aggregating multi-frame traffic network information to help predict the semaphore of each node of the traffic network. The method realizes better performance while having medium computational complexity, and has obvious advantages in practical application.
本发明公开了一种城市道路交通流量感知预测方法及其系统,系统包括数据层、转换层、表征层,其中,数据层位于交通流量感知预测系统框架的底层,转换层是沟通数据层和表征层的中介层,表征层位于系统框架顶层。表征层模型利用交通信号时空序列数据进行自监督的预训练,提取交通路网监测信号的时空特征。多层自注意力交通预测模型包括:1)道路网络空间编码器,该模块捕获单帧交通路网的空间信息;2)交通序列时间编码器,该模块捕获多帧交通路网的时间相关性;3)特征聚合器,聚合多帧交通路网信息,以帮助预测交通路网各节点的信号量。该方法在具有中等计算复杂度的同时实现了更好的性能,在实际应用中优势明显。
Urban road traffic flow perception prediction method and system
一种城市道路交通流量感知预测方法及其系统
2024-07-26
Patent
Electronic Resource
Chinese
Traffic flow prediction method and system for urban road network
European Patent Office | 2021
|Urban traffic road network traffic flow prediction method considering carbon emission model
European Patent Office | 2024
|