城市轨道交通网络短时OD客流多任务协同预测

doi:10.16097/j.cnki.1009-6744.2025.05.023

交通运输系统工程与信息 ›› 2025, Vol. 25 ›› Issue (5): 261-270.DOI: 10.16097/j.cnki.1009-6744.2025.05.023

城市轨道交通网络短时OD客流多任务协同预测

杨静^*1，侯宇晴¹，谢余晨¹，杨安安²

1. 北京建筑大学，土木与交通工程学院，北京100044；2.北京市智慧交通发展中心，北京101117

收稿日期:2025-05-16 修回日期:2025-06-26 接受日期:2025-07-07 出版日期:2025-10-25 发布日期:2025-10-25
作者简介:杨静(1980—)，女，河北张家口人，副教授，博士。
基金资助:
国家自然科学基金(52402377)；北京市自然科学基金(8252005)。

Multi-task Cooperative Prediction of Short-term Origin-Destination Flows in Urban Rail Transit Networks

YANG Jing^*1, HOU Yuqing¹, XIE Yuchen¹, YANG Anan²

1. School of Civil and Transportation Engineering, Beijing University of Civil Engineering and Architecture, Beijing 100044, China; 2. Beijing Intelligent Transportation Development Center, Beijing 101117, China

Received:2025-05-16 Revised:2025-06-26 Accepted:2025-07-07 Online:2025-10-25 Published:2025-10-25
Supported by:
National Natural Science Foundation of China (52402377)； Natural Science Foundation of Beijing, China (8252005)。

摘要/Abstract

摘要： 针对现有短时OD(Origin-Destination)预测方法中，OD客流与进出站客流关联性建模不足以及传统GNN(Graph Neural Network)表征能力受限的问题，本文提出一种融合多任务学习(Multi-task Learning, MTL)的时空图神经网络模型(MSTGN)。该模型将OD客流预测作为主任务，进出站流量作为辅助任务，借助消息传递图神经网络(MPNN)构建边-节点协同的时空传播模块，分别建模OD流(边特征)与站点客流(节点特征)，实现对复杂时空耦合关系的高效刻画。同时，引入任务一致性约束与动态权重机制，优化多任务协同效果。在北京市和杭州市的地铁实证数据集上进行对比实验与消融分析，结果表明：MSTGN在平均绝对误差(MAE)、均方根误差(RMSE)、加权平均绝对百分比误差(WMAPE)这3项指标上分别较次优模型HIAM(Heterogeneous Information Aggregation Model)提升约5.5%~6.0%，验证了模型的跨城市泛化能力；其中，MPNN模块贡献最大，消融后性能下降达6.12%。研究结果表明，MSTGN在提升短时OD预测精度与系统适应性方面具备显著优势，为城市轨道交通的智能调度与资源优化提供了技术支撑。

关键词: 智能交通, 短时OD预测, 多任务学习, 城市轨道交通, 时空特征

Abstract: Aiming to address the insufficient modeling of the correlation between OD flows and station-level inbound and outbound flows, this paper proposes a spatio-temporal graph neural network model (MSTGN) incorporating multi-task learning (MTL) to overcome the limitation of representation capability while using the traditional GNNs in existing short-term OD prediction methods. MSTGN takes the prediction of OD flow as the main task and the prediction of station passenger flow as the auxiliary task, leveraging a message-passing graph neural network (MPNN) to build an edge-node collaborative spatio-temporal propagation module that models OD flows (edge features) and station flows (node features) separately, thus enabling efficient characterization of complex spatio-temporal coupling patterns. Moreover, a task consistency constraint and a dynamic weighting mechanism are introduced to enhance a multi-task collaboration. The comparative experiments and ablation studies conducted on empirical datasets from the subway systems in Beijing and Hangzhou demonstrate that MSTGN achieves an improvement of approximately 5.5%~6.0% in MAE, RMSE, and WMAPE compared with the suboptimal model HIAM, verifying the cross-city generalization ability of this model. The MPNN module shows the most significant contribution, with a performance degradation of 6.12% upon ablation. These results indicate that MSTGN offers notable advantages in improving the accuracy and system adaptability of short-term OD prediction, providing a strong technical support for intelligent scheduling and resource optimization in urban rail transit systems.

Key words: intelligent transportation, short-term OD prediction, multi-task learning, urban rail transit, spatio-temporal features

中图分类号:

U293.13

杨静, 侯宇晴, 谢余晨, 杨安安. 城市轨道交通网络短时OD客流多任务协同预测[J]. 交通运输系统工程与信息, 2025, 25(5): 261-270.

YANG Jing, HOU Yuqing, XIE Yuchen, YANG Anan. Multi-task Cooperative Prediction of Short-term Origin-Destination Flows in Urban Rail Transit Networks[J]. Journal of Transportation Systems Engineering and Information Technology, 2025, 25(5): 261-270.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: http://www.tseit.org.cn/CN/10.16097/j.cnki.1009-6744.2025.05.023

http://www.tseit.org.cn/CN/Y2025/V25/I5/261

参考文献

[1] 张人杰,叶茂,金旭,等.基于时空特征的城市轨道交通短时OD估计方法[J].南京理工大学学报, 2024, 48(4): 512-519. [ZHANG R J, YE M, JIN X, et al. Short-term OD estimation for urban rail transit based on spatio-temporal characteristics[J]. Journal of Nanjing University of Science and Technology, 2024, 48(4): 512-519.]

[2] YAO X M, ZHAO P, YU D D. Real-time origin-destination matrices estimation for urban rail transit network based on structural state-space model[J]. Journal of Central South University, 2015, 22(11): 4498-4506.

[3] 周雨阳,李世堃,胡世龙,等.基于张量分解的城市轨道出行模式归因分析[J].铁路科学与工程学报, 2025, 22(5): 2000-2012. [ZHOU Y Y, LI S K, HU S L, et al. Attribution analysis on the travel mode of urban rail based on tensor decomposition[J]. Journal of Railway Science and Engineering, 2025, 22(5): 2000-2012.

[4] CHENG Z H, TRÉPANIER M, SUN L J. Real-time forecasting of metro origin-destination matrices with high-order weighted dynamic mode decomposition[J]. Transportation Science, 2022, 56(4): 904-918.

[5] WAND J, ZHANG Y, WEI Y, et al. Metro passenger flow prediction via dynamic hypergraph convolution networks[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 22(12): 7891-7903.

[6] LI Z, WANG X, CAI H, et al. Novel hybrid spatiotemporal convolution neural network model for short-term passenger flow prediction in a large-scale metro system[J]. Journal of Transportation Engineering, Part A: Systems, 2024, 150(5): 04024016.

[7] ZHAO L, et al. T-GCN: A temporal graph convolutional network for traffic prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(9): 3848-3858.

[8] 杨越迪,潘保霏,刘军,等.多分辨率多粒度时空特征提取的轨道交通短时OD客流预测方法[J]. 铁道学报, 2024, 46(11): 1-11. [YANG Y D, PAN B F, LIU J, et al. A short-term OD passenger flow prediction method for rail transit based on multi-resolution and multi-granularity spatio-temporal feature extraction[J]. Journal of the China Railway Society, 2024, 46(11): 1-11.]

[9] YU B, YIN H, ZHU Z. Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting[J]. arXiv Preprint arXiv: 1709.04875, 2017.

[10] GUO K, HU Y, QIAN Z, et al. Dynamic graph convolution network for traffic forecasting based on latent network of laplace matrix estimation[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 23(2): 1009-1018.

[11] XU M, DAI W, LIU C, et al. Spatial-temporal transformer networks for traffic flow forecasting[J]. arXiv Preprint arXiv: 2001.02908, 2020.

[12] WANG, Y, YIN, H, CHEN, H, et al. Origin-destination matrix prediction via graph convolution: A new perspective of passenger demand modeling[C]//Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019: 1227-1235.

[13] SHI H, YAO Q, GUO Q, et al. Predicting origin-destination flow via multi-perspective graph convolutional network[C]//2020 IEEE 36th International Conference on Data Engineering, IEEE, 2020: 1818-1821.

[14] LIU L, ZHU Y, LI G, et al. Online metro origin-destination prediction via heterogeneous information aggregation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(3): 3574-3589.

[1]	田大新, 肖啸, 周建山. AI驱动的自动驾驶汽车轨迹预测方法综述[J]. 交通运输系统工程与信息, 2025, 25(5): 1-24.
[2]	程国柱, 王文志, 杨子涵, 汪国鹏, 陈永胜, 顾爽. 网联自动驾驶车辆专用车道设置研究综述[J]. 交通运输系统工程与信息, 2025, 25(5): 25-39.
[3]	吴付威, 杨慧, 马勇, 崔同川, 闫晟煜, 张智. L3级自动驾驶车辆两阶段预警对接管绩效影响研究[J]. 交通运输系统工程与信息, 2025, 25(5): 40-49.
[4]	窦雪萍, 杨明慧, 熊杰. 考虑自适应控制点的自动驾驶公交运行优化方法[J]. 交通运输系统工程与信息, 2025, 25(5): 50-58.
[5]	宋浪, 胡晓伟, 俞山川, 安实. 网联自动驾驶环境下交叉口车道分配与车辆轨迹协同优化[J]. 交通运输系统工程与信息, 2025, 25(5): 59-71.
[6]	王殿海, 王奕飞, 黄宇浪, 刘泳, 曾佳棋. 车牌识别设备稀疏布局下的路段行程时间估计方法[J]. 交通运输系统工程与信息, 2025, 25(5): 83-90.
[7]	林培群, 黄超铄, 周楚昊, 庞崇浩, 邓锴宇. 数据驱动的高速公路自学习元胞传输模型[J]. 交通运输系统工程与信息, 2025, 25(5): 103-113.
[8]	李传耀, 章孟丹, 唐铁桥. 考虑模块化公交供应限制的双模式早通勤均衡分析[J]. 交通运输系统工程与信息, 2025, 25(5): 205-214.
[9]	沈鹏举, 宋丽英, 林川倩. 高铁列车晚点数据的异常检测与增量聚类方法[J]. 交通运输系统工程与信息, 2025, 25(5): 236-247.
[10]	孟冉, 杨洋, 宋子龙. 城市轨道交通分时定价策略的公益性水平评价[J]. 交通运输系统工程与信息, 2025, 25(5): 248-260.
[11]	姚振兴, 刘贤, 赵一飞, 王亮, 王彦琛. 手机信令不均匀定位下出行端点自适应识别方法[J]. 交通运输系统工程与信息, 2025, 25(4): 44-52.
[12]	张鹏, 李兴旺, 姬炳豪, 孙超, 李文权. 路口重复放行的公交与社会车辆协同绿波优化模型[J]. 交通运输系统工程与信息, 2025, 25(4): 53-62.
[13]	代亮, 杜鹏飞, 黄自彬, 杨朋博. 基于深度强化学习的城市交通信号分层协同控制方法[J]. 交通运输系统工程与信息, 2025, 25(4): 63-72.
[14]	王庞伟, 王思淼, 雷方舒, 徐京辉, 王子鹏, 王力. 混合动作表示强化学习下的城市交叉口智慧信控方法[J]. 交通运输系统工程与信息, 2025, 25(4): 73-83.
[15]	王连震, 沈超文, 王宇萍, 薛淑祺. 网联高速公路合流区基于间隙优化的车辆协同控制方法[J]. 交通运输系统工程与信息, 2025, 25(4): 84-95.

城市轨道交通网络短时OD客流多任务协同预测

Multi-task Cooperative Prediction of Short-term Origin-Destination Flows in Urban Rail Transit Networks

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics