Predicting Citywide Road Traffic Flow Using Deep Spatiotemporal Neural Networks

Jia, Tao; Yan, Penggao

doi:10.1109/tits.2020.2979634

Cited by 44 publications

(11 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Representative traffic flow patterns were constructed based on traffic volume and speed data in existing studies as follows. Jia et al [ 34 ] proposed a spatiotemporal neural network model to predict traffic flow for each road segment. The traffic flow was divided into recent, daily and weekly parts.…”

Section: Literature Reviewmentioning

confidence: 99%

A Novel Environment Estimation Method of Whole Sample Traffic Flows and Emissions Based on Multifactor MFD

Zang

Jiao

Song

et al. 2022

IJERPH

View full text Add to dashboard Cite

Vehicle emissions seriously affect the air environment and public health. The dynamic estimation method of vehicle emissions changing over time on the road network has always been the bottleneck of air quality simulation. The dynamic traffic volume is one of the important parameters to estimate vehicle emission, which is difficult to obtain effectively. A novel estimation method of whole sample traffic volumes and emissions on the entire road network based on multifactor Macroscopic Fundamental Diagram (MFD) is proposed in this paper. First, the intelligent clustering and recognition methods of traffic flow patterns are constructed based on neural network and deep-learning algorithms. Then, multifactor MFD models are developed considering different road types, traffic flow patterns and weekday peak hours. Finally, the high spatiotemporal resolution estimation method of whole sample traffic volumes and emissions are constructed based on MFD models. The results show that traffic flow patterns are clustered efficiently by the Self-Organizing Maps (SOM) algorithm combined with the direct time-varying speed index, which describe 91.7% traffic flow states of urban roads. The Deep Belief Network (DBN) algorithm precisely recognizes 92.1% of the traffic patterns based on the speeds of peak hours. Multifactor MFD models estimate the whole sample traffic volumes with a high accuracy of 91.6%. The case study shows that the vehicle emissions are evaluated dynamically based on the novel estimation method proposed in this paper, which is conducive to the coordinated treatment of air pollution.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

A Novel Environment Estimation Method of Whole Sample Traffic Flows and Emissions Based on Multifactor MFD

Zang

Jiao

Song

et al. 2022

IJERPH

View full text Add to dashboard Cite

show abstract

“…Driving condition is a dynamic circumstances that a vehicle is running along the pathway [10]. It includes vehicle motion state variables such like second-by-second velocity , acceleration , VSP [11] , time step, spatial context variables such as location and slope, and ancillary state variables such as energy category and whether passengers are carried.…”

Section: Data Preliminarymentioning

confidence: 99%

Dynamic Capture and Estimation of Energy-Emission Benefits for Electric Taxicabs with Spatiotemporal Big Data and Deep-Learning-Based Microscopic Model

Chen

Liu

Jia

2022

Advances in Transdisciplinary Engineering

View full text Add to dashboard Cite

Reducing carbon emissions to cope with climate change and short of energy have become a global trend, so it is urgent to accurately measure energy consumption and emissions. As taxi occupies the second highest proportion of domestic roads, it is necessary to study the emission of new energy. However, existing studies often consider the whole region, with low accuracy and no true value, resulting in difficult verification of conclusions. This paper proposed a micro-energy consumption and carbon emission model for taxicabs based on trajectory data and deep learning method to dynamically simulate the real-time energy consumption of taxicabs with different energy sources. First, this paper detected the correlation between driving state and energy consumption carbon emission in portable emissions measurement system (PEMS) environment. Then, a deep learning-based framework was built to learn the vehicle’s energy consumption carbon emission pattern. In particular, Gated Recurrent Unit (GRU) neural network is used to learn current and historical driving habits and the influence of external environment on energy consumption, while life cycle assessment (LCA) method is used to obtain the emission patterns of vehicles with different energy types in the whole life cycle. The measured data are obtained in Wuhan, the precision of our model is higher than that of the existing model. At the same time, we also applied it to the taxicab monthly trajectory dataset to obtain the spatial-temporal energy consumption emission patterns of different types of energy. The results show that pure electric vehicle (PEV) has obvious greenhouse gas emission reduction effect, compared with gasoline and compressed natural gas (CNG) vehicles, the emission reduction is 12.03% and 12.07% respectively, but the total energy consumption achieves little advantage. This model will lay a foundation for the formulation of regional road network emission inventory, so as to provide support for the government to make relevant decisions.

show abstract

“…Images are also data representations that can be processed by Deep Learning architectures. Some studies arrange snapshots of network-wide traffic congestion maps as a time series, and resort to Deep Learning architectures for motion prediction to estimate the future trajectory of objects [55], [193]. Other works convert traffic speed time series from multiple points of a traffic network into a heatmap, where color expresses the speed value [125], [155].…”

Section: G What Possibilities Does Data Fusion Offer?mentioning

confidence: 99%

Deep Learning for Road Traffic Forecasting: Does it Make a Difference?

Manibardo¹,

Laña²,

Ser³

2020

Preprint

View full text Add to dashboard Cite

Deep Learning methods have been proven to be flexible to model complex phenomena. This has also been the case of Intelligent Transportation Systems (ITS), in which several areas such as vehicular perception and traffic analysis have widely embraced Deep Learning as a core modeling technology. Particularly in short-term traffic forecasting, the capability of Deep Learning to deliver good results has generated a prevalent inertia towards using Deep Learning models, without examining in depth their benefits and downsides. This paper focuses on critically analyzing the state of the art in what refers to the use of Deep Learning for this particular ITS research area. To this end, we elaborate on the findings distilled from a review of publications from recent years, based on two taxonomic criteria. A posterior critical analysis is held to formulate questions and trigger a necessary debate about the issues of Deep Learning for traffic forecasting. The study is completed with a benchmark of diverse short-term traffic forecasting methods over traffic datasets of different nature, aimed to cover a wide spectrum of possible scenarios. Our experimentation reveals that Deep Learning could not be the best modeling technique for every case, which unveils some caveats unconsidered to date that should be addressed by the community in prospective studies. These insights reveal new challenges and research opportunities in road traffic forecasting, which are enumerated and discussed thoroughly, with the intention of inspiring and guiding future research efforts in this field.

show abstract

Predicting Citywide Road Traffic Flow Using Deep Spatiotemporal Neural Networks

Cited by 44 publications

References 29 publications

A Novel Environment Estimation Method of Whole Sample Traffic Flows and Emissions Based on Multifactor MFD

A Novel Environment Estimation Method of Whole Sample Traffic Flows and Emissions Based on Multifactor MFD

Dynamic Capture and Estimation of Energy-Emission Benefits for Electric Taxicabs with Spatiotemporal Big Data and Deep-Learning-Based Microscopic Model

Deep Learning for Road Traffic Forecasting: Does it Make a Difference?

Contact Info

Product

Resources

About