Search

In urban system management and public safety, accurate traffic flow forecasting is pivotal for real-world applications such as traffic control, resource-sharing scheduling platforms, and intelligent transportation systems. The challenge lies in effectively capturing complex spatio-temporal correlations. To address this, we introduce ST-DSTN, a novel spatial-temporal fusion attention-based deep neural network that significantly advances the current state-of-the-art. ST-DSTN utilizes three distinct temporal branches to capture temporal dependencies and a convolutional neural network to encapsulate spatial dependencies. Our innovative contribution includes a spatio-temporal fusion attention module, effectively modeling the dynamic spatio-temporal dependencies of traffic data, thus improving upon existing methods. Additionally, we propose a hierarchical fusion method that permeates the entire model, capturing complex correlations at various levels through multi-level fusion. Our comprehensive experiments on the BikeNYC dataset demonstrate ST-DSTN's superior performance, supported by numerical results that substantiate its improvement over current methods, surpassing the DeepSTN+ model by 4.31%. [ABSTRACT FROM AUTHOR]

In this study, a traffic queue model is employed to simulate traffic parameters for optimization purposes. The primary objective is to enhance the average queue length, cumulative delay, and cumulative reward for varying numbers of vehicles generated using an innovative approach. The proposed methodology integrates deep Reinforcement Q-Learning techniques implemented through a realistic simulator, Simulation of Urban MObility, based on OpenStreetMap. This simulator is used to balance the flow of traffic by monitoring traffic conditions and adjusting signals (Red/Yellow/Green) accordingly. The research extensively evaluates simulation outcomes across different scenarios involving 50, 100, 500, and 1000 vehicles, with episodes 10, 50 and 100. The results provide valuable insights into training times, simulation durations, and total rewards obtained, indicating improvements in mitigating traffic congestion. The introduction of a traffic control strategy enhances model effectiveness through the utilization of four-layer deep neural network architectures with 400 neurons in hidden layers. However, the study also highlights persistent challenges related to real-time implementation and complexities in state-action-agent relationships. The progress highlights improvements in reducing congestion and enhancing efficiency at junctions by dynamically adjusting signal phases while leveraging historical data. The results highlight the potential of dynamic traffic management strategies in effectively addressing urban traffic challenges. [ABSTRACT FROM AUTHOR]

Traffic flow forecasting is of great significance to urban traffic control and public safety applications. The key challenge of traffic flow forecasting is how to capture the complex correlation of different time levels and learn time dependence. Some external information is closely related to traffic flow, such as accidental traffic accidents, weather, and Point of Interests (PoI) information. This paper proposes a deep learning-based model, called AttDeepSTN+, which is used to predict the inflow and outflow of each area of the entire city. Specifically, AttDeepSTN+ uses the structure of interactive attention and convolution to model the temporal closeness, trend, and periodicity of crowd flow, in the interactive attention layer, learn the importance of closeness to periodicity and trend respectively to model the long-term dependence of time, and then use feature fusion to capture complex correlations at different levels, thereby reducing model prediction accuracy. In addition, PoI information are combined with time factors to express the influence of location attributes on crowd flow, to learn prior knowledge of crowd flow. Finally, a new fusion mechanism is used to fuse the attention layer modules and PoI information and other information together into the module to capture the complex correlation between different levels of features, to predict the final crowd flow in each area, and further improve the prediction accuracy of the model. The New York City crowd flow experiment shows that the model is better than the current state-of-the-art baseline. [ABSTRACT FROM AUTHOR]

Traffic flow forecasting or prediction plays an important role in the traffic control and management of a city. Existing works mostly train a model using the traffic flow data of a city and then test the trained model using the data of the same city. It may not be truly intelligent as there are many cities around us and there should be some shared knowledge among different cities. The data of a city and its knowledge can be used to help improve the traffic flow forecasting of other cities. To address this motivation, we study building a universal deep learning model for multi-city traffic flow forecasting. In this paper, we exploit spatial-temporal correlations among different cities with multi-task learning to approach the traffic flow forecasting tasks of multiple cities. As a result, we propose a M ulti-city T raffic flow forecasting N etwork (MTN) via multi-task learning to extract the spatial dependency and temporal regularity among multiple cities later used to improve the performance of each individual city traffic flow forecasting collaboratively. In brief, the proposed model is a quartet of methods: (1) It integrates three temporal intervals and formulates a multi-interval component for each city to extract temporal features of each city; (2) A spatial-temporal attention layer with 3D Convolutional kernels is plugged into the neural networks to learn spatial-temporal relationship; (3) As traffic peak distributions of different cities are often similar, it proposes to use a peak zoom network to learn the peak effect of multiple cities and enhance the prediction performance on important time steps in different cities; (4) It uses a fusion layer to merge the outputs from distinct temporal intervals for the final forecasting results. Experimental results using real-world datasets from DIDI show the superior performance of the proposed model. [ABSTRACT FROM AUTHOR]

Development of speed model is a challenging task particularly under mixed traffic condition where the traffic composition plays a significant role in determining the individual vehicle speed. Estimation of passenger car equivalent (PCE) which is essential in converting the mixed traffic volume into its equivalent homogeneous also requires the speed information of individual vehicle categories at varying traffic conditions on the road. The present research was carried out to model the individual vehicle speed and to study the effects of traffic volume and its composition on individual speed and PCE in the context of urban mixed traffic. Traffic data on classified traffic volume and speed information were collected at six-lane divided arterial mid-block road sections in New Delhi, India. The methodology of artificial neural network was adopted to develop a volume-based speed prediction model for individual vehicle category. Validation results showed a great deal of agreement between the predicted and the observed speeds. Then, the sensitivity analysis was performed utilizing the model developed in order to examine the effects of traffic volume and its composition on individual speed and corresponding PCE. [ABSTRACT FROM AUTHOR]

In order to ensure that traffic capacity is stable, and avoid incident congestion is anabatic, a double-layer ramp-metering model is proposed in this paper to control the traffic flow at each on-ramp, nearby incident congestion. The function of the lower model is to recognize where the incident congestion exists, based on an adaptive neural network, whose inputs are traffic flow, velocity and density. The outputs of the lower model are the section number where the congestion occurs and the ramp number which should be controlled. If there is congestion, the upper model would be activated. The function of the upper model is to calculate the control strategy. The required control ramp number from the lower model would be transmitted to the upper model together with the real-time state of controlled section. The upper model would calculate the ramp-metering ratio for each required ramp by minimizing the main time cost, total average queue length and equity index. In order to describe the traffic state more accurately, a tailored traffic model for urban freeway in China is established in the upper model to get the prediction, which considers the speed-limited value on main roadway, and corrects the calculation method of traffic flow state on origin point and exit ramp. The simulation results show the double model is effective to control the incident congestion induced by aggressive behavior. [ABSTRACT FROM AUTHOR]