基于数字孪生的城市交通流量可视预测研究
City Traffic Flow Visual Prediction Based on Digital Twin
王健松 1李学俊 1王桂娟 1郭皓 1吴亚东2
作者信息
- 1. 西南科技大学 计算机科学与技术学院,四川 绵阳 621010
- 2. 四川轻化工大学 计算机科学与工程学院,四川 自贡 645002
- 折叠
摘要
由于城市路网数据的复杂性和动态性,直接解读各道路关联关系变得困难,直接连接性的不确定性也影响了预测准确性.为解决这些问题,利用出租车轨迹数据,结合图卷积神经网络,提出了数字孪生基础上的城市交通流量智能可视预测框架.为提高预测精度,根据历史流量数据创建了路网的时空关联图,构建了时空图卷积网络的ASTRG-GCN交通流量预测模型.通过数字孪生技术,将动态交通数据和虚拟三维交通场景融合,实时模拟交通场景,为城市交通优化提供决策支持.最终,设计并实现了城市交通流量可视分析框架,使用户能够高效分析交通运行态势.实验结果表明,该模型预测精度在两个数据集上高于对比算法.数字孪生的可视分析系统实现了交通拥堵识别、交通场景模拟和交通变化对比等效果,为交通规划者提供了决策支持.
Abstract
Due to the complexity and dynamic nature of urban road network data,it is difficult to interpret the road correlation directly,and the uncertainty of direct connectivity also affects the prediction accuracy.To solve these problems,with taxi trajectory data and graph convolutional neural network,we propose an intelligent visual prediction framework for urban traffic flow based on digital twins.In order to improve the prediction accuracy,we create the spatio-temporal correlation graph of the road network based on the historical traffic data,and construct the ASTRG-GCN traffic flow prediction model of the spatio-temporal convolutional network.Through digital twin technology,we integrate dynamic traffic data and virtual three-dimensional traffic scenes to simulate traffic scenes in real time and provide decision support for urban traffic optimization.Finally,we design and implement the visual analysis framework of urban traffic flow,which enables users to analyze traffic operation situation efficiently.Experimental results show that the prediction accuracy of the proposed model is higher than that of the comparison algorithm on the two data sets.The visual analysis system of digital twin can realize the effect of traffic congestion identification,traffic scene simulation and traffic change comparison,and provide decision support for traffic planners.
关键词
数字孪生/城市交通/轨迹数据/流量预测/可视化分析Key words
digital twin/city traffic/trajectory data/flow prediction/visualization analysis引用本文复制引用
基金项目
四川省科技项目(2023YFG0307)
四川轻化工大学人才引进项目(2020RC20)
出版年
2024
国家科技期刊平台
NSTL
万方数据