城市交通路网动态短时推理与精准预测研究

doi:10.11959/j.issn.2096-6652.202233

智能科学与技术学报 ›› 2022, Vol. 4 ›› Issue (3): 380-395.doi: 10.11959/j.issn.2096-6652.202233

城市交通路网动态短时推理与精准预测研究

覃缘琪¹, 季青原², 葛俊¹, 戴星原³, 陈圆圆³, 王晓³^,⁴

¹ 之江实验室，浙江杭州 310000
² 银江技术股份有限公司，浙江杭州 310030
³ 中国科学院自动化研究所复杂系统管理与控制国家重点实验室，北京 100190
⁴ 青岛智能产业技术研究院，山东青岛 266000

修回日期:2022-07-10 出版日期:2022-09-15 发布日期:2022-09-01
作者简介:覃缘琪（1992- ），女，之江实验室智能社会治理研究中心研究专员，主要研究方向为智能交通控制、智能社会治理等
季青原（1989- ），男，银江技术股份有限公司博士后，主要研究方向为交通预测、图神经网络、知识图谱
葛俊（1989- ），男，之江实验室智能社会治理研究中心高级研究专员，主要研究方向为强化学习、群体智能
戴星原（1993- ），男，中国科学院自动化研究所复杂系统管理与控制国家重点实验室博士生，主要研究方向为智能交通系统、机器学习和深度学习
陈圆圆（1989- ），男，博士，中国科学院自动化研究所复杂系统管理与控制国家重点实验室副研究员，主要研究方向为社会交通、数据驱动的交通建模与预测
王晓（1988- ），女，博士，中国科学院自动化研究所复杂系统管理与控制国家重点实验室副研究员，主要研究方向为群体行为的激发与汇聚激励、群体智能和社交网络挖掘与分析
基金资助:
国家重点研发计划基金资助项目(SQ2019YFE012476)

Short-term traffic state reasoning and precise prediction in urban networks

Yuanqi QIN¹, Qingyuan JI², Jun GE¹, Xingyuan DAI³, Yuanyuan CHEN³, Xiao WANG³^,⁴

¹ Zhejiang Lab, Hangzhou 310000, China
² Enjoyor Technology Co., Ltd., Hangzhou 310030, China
³ The State Key Laboratory for Management and Control of Complex Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China
⁴ Qingdao Academy of Intelligent Industries, Qingdao 266000, China

Revised:2022-07-10 Online:2022-09-15 Published:2022-09-01
Supported by:
The National Key Research and Development Program of China(SQ2019YFE012476)

摘要/Abstract

摘要：

城市路网结构对于交通拥堵的形成及其在时空上的传播过程具有重要的影响。然而，在基于传统交通模型或深度学习模型的研究中，交通模式的生成与传播往往并未考虑路网特征，且只能通过交通特征指标进行间接刻画，使得难以动态描述其在时间及空间维度上的特征，导致交通模式传播预测精准度不高且缺乏针对性。为了解决上述问题，提出了一种新型的基于交通模式推理的交通预测框架 TP²。该框架将拥堵传播模式建模为一个随时间变化的时序知识图谱，并使用一种包含了全新聚合函数RGraAN的推理框架进行时序推理，以捕捉交通拥堵的动态时变传播模式，将路段和与其存在交通模式关联的路段进行组合，并构建时空关联路网子区域，然后基于图神经网络的交通短时预测模块充分挖掘子区内的交通流时空相关性，并预测子区内各个路段的未来速度变化情况。与现有方法相比，TP²预测精度相比A3T-GCN模型有1%~2%的提升。

关键词: 路网结构, 传播模式, 交通状态推理与预测, 时序知识图谱, 图神经网络

Abstract:

The structure of urban traffic network has a significant impact on the formation and spatio-temporal pattern propagation of traffic congestions.However, in studies based on traditional traffic models or deep learning models, the generation of traffic mode can only be described indirectly by traffic indicators, without considering the traffic network feature.This makes it very difficult to accurately describe the propagation dynamics both in temporal and spatial dimensions and lacks specificity.To tackle the above-mentioned problems, a novel traffic state prediction approach based on traffic pattern reasoning (TP²) framework was proposed.The framework modeled congestion propagation as a dynamically evolving temporal knowledge graph (TKG), and applied an inferencing framework (TPP-TKG) that was based on a novel aggregator called RGraAN.TPP-TKG captured the spatial-temporal propagation pattern of traffic congestion, and combined related road links to a given link, and constructed correlated sub region of the traffic network.Then a traffic state predicting based on graph neural network was employed to predict short-term speed evolution of road links in this sub region.Comparing to the state-of-the-art benchmark models, TP² achieves 1% ~ 2% higher accuracy.

Key words: traffic network structure, propagation mode, traffic mode reasoning and prediction, temporal knowledge graph, graph neural network.

中图分类号:

U495

覃缘琪, 季青原, 葛俊, 等. 城市交通路网动态短时推理与精准预测研究[J]. 智能科学与技术学报, 2022, 4(3): 380-395.

Yuanqi QIN, Qingyuan JI, Jun GE, et al. Short-term traffic state reasoning and precise prediction in urban networks[J]. Chinese Journal of Intelligent Science and Technology, 2022, 4(3): 380-395.

图/表 18

图1

图2

图3

图4

图5

图6

表1

表2

表3

表4

表5

TP2与其他基准模型在早高峰与晚高峰时段上的时序预测表现"

区域	类别	模型	早高峰				晚高峰
区域	类别	模型	Hit@1	Hit@3	Hit@10	MRR	Hit@1	Hit@3	Hit@10	MRR
区域1	非知识图	HA	13.2				11.6
	谱模型	ARIMA	18.6				16.4
		ST-GCN	19.6				17.5
		AST-GCN	19.4				20.3
	知识图谱	T-TransE	28.0	55.3	69.3	35.1	25.1	39.7	58.2	32.2
	模型	TA-TransE	35.6	52.7	76.7	48.9	23.9	62.1	75	38.6
		TA-DistMult	57.1	61.8	83.1	57.5	41.0	65.1	72.8	52.8
		RE-NET	64.3	71.5	85.8	72.1	51.2	75.6	81.9	58.7
		TP²	$66 . 1$	$72 . 9$	$87 . 2$	$74 . 3$	$52 . 9$	$77 . 3$	$83 . 6$	$60 . 2$
区域2	非知识图	HA	11.2				14.2
	谱模型	ARIMA	14.5				16.6
		ST-GCN	16.7				18.5
		AST-GCN	22.7				21.1
	知识图谱	T-TransE	32.5	32.5	47.8	24.3	27.2	45.7	66.2	33.4
	模型	TA-TransE	29.6	47.7	51.3	41.3	28.8	53.3	69.1	48.9
		TA-DistMult	42.5	58.4	73.5	52.8	43.0	57.9	76.3	57.9
		RE-NET	57.1	68.7	78.4	62.4	56.5	70.3	84.8	69.6
		TP²	$58 . 9$	$68 . 5$	$80 . 1$	$65 . 3$	$57 . 7$	$74 . 4$	$86 . 7$	$71 . 8$

表5

表7

图7

表6

表8

表9

图8

图9

参考文献 46

[47]	LIU K H , YE Z H , GUO H Y ,et al. FISS GAN:a generative adversarial network for foggy image semantic segmentation[J]. IEEE/CAA Journal of Automatica Sinica, 2021,8(8): 1428-1439.
[48]	SIMS A G , DOBINSON K W . The Sydney coordinated adaptive traffic (SCAT) system philosophy and benefits[J]. IEEE Transactions on Vehicular Technology, 1980,29(2): 130-137.
[49]	CHEN Z H , YANG Y J , HUANG L P ,et al. Discovering urban traffic congestion propagation patterns with taxi trajectory data[J]. IEEE Access, 2018,6: 69481-69491.
[50]	YU B , YIN H T , ZHU Z X . Spatio-temporal graph convolutional networks:a deep learning framework for traffic forecasting[C]// Proceedings of the 27th International Joint Conference on Artificial Intelligence. California:International Joint Conferences on Artificial Intelligence Organization, 2018.
[51]	GUO S N , LIN Y F , FENG N ,et al. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019,33: 922-929.
[52]	ADLER J , LUNZ S . Banach wasserstein GAN[J]. arxiv preprint, 2018,arXiv:1806.06621.
[1]	WANG H H , CAO R X , ZENG W H . Multi-agent based and system dynamics models integrated simulation of urban commuting relevant carbon dioxide emission reduction policy in China[J]. Journal of Cleaner Production, 2020,272:122620.
[2]	CHEN S H , ZHU F H , WANG F Y . An erlang-based simulation approach of artificial transportation systems[C]// Proceedings of 2016 IEEE 19th International Conference on Intelligent Transportation Systems. Piscataway:IEEE Press, 2016: 24-28.
[3]	JIN J C , MA X L . A non-parametric Bayesian framework for traffic-state estimation at signalized intersections[J]. Information Sciences, 2019,498: 21-40.
[4]	WU J P , CAO M , CHEUNG J C K ,et al. TeMP:temporal message passing for temporal knowledge graph completion[J]. arXiv preprint,2020,arXiv:2010.03526.
[5]	SUN S M , CHEN J , SUN J . Traffic congestion prediction based on GPS trajectory data[J]. International Journal of Distributed Sensor Networks, 2019,15(5): 155014771984744.
[6]	贺冬冬 . 基于出租车轨迹数据的短时交通流预测模型研究[D]. 大连:大连海事大学, 2013.
	HE D D . The short-time prediction model of traffic flow based on taxi tragectory data[D]. Dalian:Dalian Maritime University, 2013.
[7]	ZHOU X Y , WANG W J , YU L . Traffic flow analysis and prediction based on GPS data of floating cars[M]// Lecture notes in electrical engineering. Heidelberg: Springer, 2012: 497-508.
[8]	FOULADGAR M , PARCHAMI M , ELMASRI R ,et al. Scalable deep traffic flow neural networks for urban traffic congestion prediction[C]// Proceedings of 2017 International Joint Conference on Neural Networks. Piscataway:IEEE Press, 2017: 2251-2258.
[9]	ZHU J W , HAN X , DENG H H ,et al. KST-GCN:a knowledge-driven spatial-temporal graph convolutional network for traffic forecasting[J]. IEEE Transactions on Intelligent Transportation Systems, 2022: 1-11.
[10]	PETTY K F , MOYLAN A J . Traffic bottleneck detection and classification on a transportation network graph:US9330565[P]. 2016-05-03.
[11]	CUI Z Y , KE R M , PU Z Y ,et al. Deep bidirectional and unidirectional LSTM recurrent neural network for network-wide traffic speed prediction[J]. arXiv preprint,2018,arXiv:1801.02143.
[12]	宋珊珊 . 基于浮动车数据的短时交通流预测研究[D]. 大连:大连理工大学, 2019.
	SONG S S . Short-term traffic flow prediction based on floating car data[D]. Dalian:Dalian University of Technology, 2019.
[13]	赵宏, 翟冬梅, 石朝辉 . 短时交通流预测模型综述[J]. 都市快轨交通, 2019,32(4): 50-54.
	ZHAO H , ZHAI D M , SHI C H . Review of short-term traffic flow forecasting models[J]. Urban Rapid Rail Transit, 2019,32(4): 50-54.
[14]	CHIABAUT N , FAITOUT R . Traffic congestion and travel time prediction based on historical congestion maps and identification of consensual days[J]. Transportation Research Part C:Emerging Technologies, 2021,124:102920.
[15]	PAN B , DEMIRYUREK U , SHAHABI C . Utilizing real-world transportation data for accurate traffic prediction[C]// Proceedings of 2012 IEEE 12th International Conference on Data Mining. Piscataway:IEEE Press, 2012: 595-604.
[16]	WILLIAMS B M , HOEL L A . Modeling and forecasting vehicular traffic flow as a seasonal ARIMA process:theoretical basis and empirical results[J]. Journal of Transportation Engineering, 2003,129(6): 664-672.
[17]	ALGHAMDI T , ELGAZZAR K , BAYOUMI M ,et al. Forecasting traffic congestion using ARIMA modeling[C]// Proceedings of 2019 15th International Wireless Communications ＆ Mobile Computing Conference. Piscataway:IEEE Press, 2019: 1227-1232.
[18]	CHEN C Y , HU J M , MENG Q ,et al. Short-time traffic flow prediction with ARIMA-GARCH model[C]// Proceedings of 2011 IEEE Intelligent Vehicles Symposium. Piscataway:IEEE Press, 2011: 607-612.
[19]	DU Q B , YIN F M , LI Z C . Base station traffic prediction using XGBoost-LSTM with feature enhancement[J]. IET Networks, 2020,9(1): 29-37.
[20]	SUN B , SUN T , JIAO P P . Spatio-temporal segmented traffic flow prediction with ANPRS data based on improved XGBoost[J]. Journal of Advanced Transportation,2021, 2021:5559562.
[21]	VLAHOGIANNI E I , KARLAFTIS M G , KOPELIAS P . Modeling freeway travel speed across lanes:a vector autoregressive approach[C]// Proceedings of the 13th International IEEE Conference on Intelligent Transportation Systems. Piscataway:IEEE Press, 2010: 569-574.
[22]	ANWAR T , LIU C F , VU H L ,et al. Capturing the spatiotemporal evolution in road traffic networks[J]. IEEE Transactions on Knowledge and Data Engineering, 2018,30(8): 1426-1439.
[23]	REMPE F , HUBER G , BOGENBERGER K . Spatio-temporal congestion patterns in urban traffic networks[J]. Transportation Research Procedia, 2016,15: 513-524.
[24]	LYU Y S , DUAN Y J , KANG W W ,et al. Traffic flow prediction with big data:a deep learning approach[J]. IEEE Transactions on Intelligent Transportation Systems, 2015,16(2): 865-873.
[25]	LI Z S , XIONG G , CHEN Y Y ,et al. A hybrid deep learning approach with GCN and LSTM for traffic flow prediction[C]// Proceedings of 2019 IEEE Intelligent Transportation Systems Conference. Piscataway:IEEE Press, 2019: 1929-1933.
[26]	CHEN Y Y , CHEN H Y , YE P J ,et al. Acting as a decision maker:traffic-condition-aware ensemble learning for traffic flow prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2022,23(4): 3190-3200.
[27]	DAI X Y , FU R , ZHAO E M ,et al. DeepTrend 2.0:a light-weighted multi-scale traffic prediction model using detrending[J]. Transportation Research Part C:Emerging Technologies, 2019,103: 142-157.
[28]	LIN Y L , DAI X Y , LI L ,et al. Pattern sensitive prediction of traffic flow based on generative adversarial framework[J]. IEEE Transactions on Intelligent Transportation Systems, 2019,20(6): 2395-2400.
[29]	ZHANG W , ZHU F H , CHEN Y Y ,et al. Differential time-variant traffic flow prediction based on deep learning[C]// Proceedings of 2020 IEEE 23rd International Conference on Intelligent Transportation Systems. Piscataway:IEEE Press, 2020: 1-6.
[30]	LIN Y L , DAI X Y , LI L ,et al. An efficient deep reinforcement learning model for urban traffic control[J]. arXiv preprint,2018,arXiv:1808.01876.
[31]	CHEN Y W , WANG X , LI L X ,et al. Traffic situational awareness research and development enhanced by social media data:the state of the art and prospects[J]. Chinese Journal of Intelligent Science and Technology, 2022,4(1): 1-14.
[32]	XU X J , GAO X B , XU Z Z ,et al. TCPModel:a short-term traffic congestion prediction model based on deep learning[M]// Communi cations in computer and information science. Singapore: Springer Singapore, 2019.
[33]	ZHANG X Y , HUANG C , XU Y ,et al. Traffic flow forecasting with spatial-temporal graph diffusion network[J]. arXiv preprint,2021,arXiv:2110.04038.
[34]	ZHU J W , WANG Q J , TAO C ,et al. AST-GCN:attribute-augmented spatiotemporal graph convolutional network for traffic forecasting[J]. IEEE Access, 2021,9: 35973-35983.
[35]	ZHENG C P , FAN X L , WANG C ,et al. GMAN:a graph multi-attention network for traffic prediction[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020,34(1): 1234-1241.
[36]	NAGY A M , SIMON V . Improving traffic prediction using congestion propagation patterns in smart cities[J]. Advanced Engineering Informatics, 2021,50:101343.
[37]	张阳, 胡月, 辛东嵘 . 一种考虑时空关联的深度学习短时交通流预测方法[J]. 智能科学与技术学报, 2021,3(2): 172-178.
	ZHANG Y , HU Y , XIN D R . A deep learning short-term traffic flow prediction method considering spatial-temporal association[J]. Chinese Journal of Intelligent Science and Technology, 2021,3(2): 172-178.
[38]	陈美林, 郑治豪, 郭宝 ,等. 基于因果关联的交通拥堵传播分析[J]. 中南大学学报(自然科学版), 2020,51(12): 3575-3583.
	CHEN M L , ZHENG Z H , GUO B ,et al. Traffic congestion spreading analysis based on causal nexus[J]. Journal of Central South University (Science and Technology), 2020,51(12): 3575-3583.
[39]	NAGY A M , SIMON V . A novel congestion propagation modeling algorithm for smart cities[J]. Pervasive and Mobile Computing, 2021,73:101387.
[40]	陆浩, 王飞跃, 刘德荣 ,等. 基于科研知识图谱的近年国内外自动化学科发展综述[J]. 自动化学报, 2014,40(5): 994-1015.
	LU H , WANG F Y , LIU D R ,et al. Analytics of lastest research progress in automation discipline based on academic knowledge mapping[J]. Acta Automatica Sinica, 2014,40(5): 994-1015.
[41]	赵学亮, 王涛, 王晓 ,等. 基于文献指标与合著网络的《自动化学报》2011-2016年发表论文分析研究[J]. 自动化学报, 2017,43(12): 2232-2243.
	ZHAO X L , WANG T , WANG X ,et al. A literature study on Acta Automatica Sinica during 2010 to 2016 with bibliographic and coauthorship analysis[J]. Acta Automatica Sinica, 2017,43(12): 2232-2243.
[42]	JIN J C , GUO H F , XU J ,et al. An end-to-end recommendation system for urban traffic controls and management under a parallel learning framework[J]. IEEE Transactions on Intelligent Transportation Systems, 2021,22(3): 1616-1626.
[43]	JIN W , QU M , JIN X S ,et al. Recurrent event network:autoregressive structure inference over temporal knowledge graphs[J]. arXiv preprint,2019,arXiv:1904.05530.
[44]	CHO K , VAN MERRIENBOER B , GULCEHRE C ,et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation[J]. arxiv preprint,2014,arXiv:1406,1078.
[45]	ZHANG Y X , WANG S Z , CHEN B ,et al. TrafficGAN:network-scale deep traffic prediction with generative adversarial nets[J]. IEEE Transactions on Intelligent Transportation Systems, 2021,22(1): 219-230.
[46]	JIN J C , RONG D D , ZHANG T ,et al. A GAN-based short-term link traffic prediction approach for urban road networks under a parallel learning framework[J]. IEEE Transactions on Intelligent Transportation Systems, 2022(99): 1-12.

路段ID	时间戳	速度/（km·h^-1）	信号周期/s
R0001	2020-01-01 08:00:00	35	180
R0001	2020-01-01 08:05:00	32	180
R0001	2020-01-01 08:10:00	28	135

拥堵率	0＜v＜10	10＜v＜20	20＜v＜40
0＜cr＜0.2	1	6	11
0.2＜cr＜0.4	2	7	12
0.4＜cr＜0.6	3	8	13
0.6＜cr＜0.8	4	9	14
0.8＜cr＜1.0	5	10	15

交通传播模式关系	具体解释
p₀	cr相同，v相同
p₁	cr相同，v更高
p₂	cr更低，v更高
p₃	cr更低，v相同
p₄	cr更低，v更低
p₅	cr相同，v更低
p₆	cr更高，v更低
p₇	cr更高，v相同
p₈	cr更高，v更高

基准模型	基准模型最佳超参
SVR	使用线性核函数，惩罚因子为0.001
GRU	使用单层GRU网络，隐藏向量维度为64
GCN	使用双层GCN，隐藏向量维度为64
T-GCN	隐藏向量维度为64
ST-GCN	ST-Conv的通道数分别为64、16、64；图卷积核数量K与时间卷积核数量K_t均设为3
A3T-GCN	隐藏向量维度为64

拥堵频率	时间窗口5 min	时间窗口10 min	时间窗口20 min
60%	67.3/55.4/58.2/66.5	70.3/58.5/64.1/69.2	70.9/57.2/61.5/68.5
80%	71.7/59.2/62.1/69.8	74.3/60.2/65.3/71.8	72.1/59.3/63.5/70.2
90%	70.4/57.2/59.3/67.5	72.1/59.8/63.7/69.7	71.2/58.1/62.5/67.7

城市交通路网动态短时推理与精准预测研究

Short-term traffic state reasoning and precise prediction in urban networks

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 18

参考文献 46

相关文章 0

Metrics

推荐阅读 0

模型	MAE	RMSE	MAPE
HA	7.61/9.14/12.0	8.87/11.4/16.5	21.8%/25.7%/33.9%
SVR	7.32/8.72/11.6	8.63/10.5/14.3	16.5%/19.8%/23.5%
ARIMA	5.18/6.79/9.27	6.02/7.11/9.74	14.8%/17.5%/20.1%
GCN	6.46/7.78/9.19	6.94/8.54/9.81	11.6%/13.2%/15.4%
GRU	5.34/5.87/7.27	6.48/6.88/8.53	11.2%/12.4%/14.0%
T-GCN	4.42/4.86/5.62	5.60/5.91/6.52	10.7%/11.5%/12.8%
ST-GCN	3.88/4.41/4.99	5.27/5.45/5.86	9.5%/9.8%/10.9%
A3T-GCN	3.65/3.89/4.24	4.62/4.93/5.34	9.2%/9.6%/10.5%
TP²	2.99/3.21/3.39	4.05/4.40/4.59	8.6%/8.9%/9.3%

模型	训练时间/s	推理时间/s
A3T-GCN	8 532	115
TP²	2 480	26