Traffic flow in a road network is mutually interactive and interdependent with each other. It is challenging to describe the dynamics of traffic network flow by using analytical methods. The deep convolutional neural network (DCNN) model is employed to address traffic network flow prediction. To improve the parameter learning efficiency confronting traffic big data, a parallel training approach is developed for the DCNN prediction model. The theoretical foundation is developed for the parallel training algorithm of the DCNN model. A master-slave parallel computing solution for traffic network flow prediction is implemented on the Spark cloud. Real data of traffic network flow is applied to verify the effectiveness of the DCNN prediction model and the parallel training algorithm. The experimental results demonstrate that the DCNN prediction model for traffic network flow outperforms the typical prediction models based on backpropagation neural networks, support vector regressions, radial basis functions, and decision tree regressions. The proposed parallel training method can improve the training efficiency and obtain global features of the entire dataset from local learning with regard to the respective data subsets.

When dealing with traffic big data under the background of Internet of Things (IoT), traffic control under the single-machine computing environment is difficult to adapt to the massive and rapid analysis and decision-making. To tackle this problem, we propose a parallel computing approach of traffic network flow control based on the mechanism of model predictive control (MPC). A non-analytical rule-based traffic flow model is developed to forecast vehicle movements in the prediction horizon according to the real-time feedback information of traffic flow, and evaluate the performances of candidate control strategies. Furthermore, to accelerate the solution process of obtaining the optimal control schemes in the prediction horizon, a two-level hierarchical parallel genetic algorithm (HPGA) based on Spark cloud computing is designed. Through the parallel computing architecture, the computationally intensive optimization tasks are decomposed into multiple parallel sub-tasks with the aid of resilient distributed datasets (RDDs), which improves the computational efficiency. The simulation results demonstrate the validity of the proposed methodology for traffic network flow predictive control with respect to unsaturated and oversaturated traffic scenarios. The Spark-based parallel optimization approach has the potential to satisfy the computing requirements of online optimization when dealing with the big data of traffic network flow control while keeps favorable control performances.

The escalating air pollution resulting from traffic congestion has necessitated a shift in traffic control strategies towards green and low-carbon objectives. In this study, a graph convolutional network and self-attention value decomposition-based multi-agent actor-critic (GSAVD-MAC) approach is proposed to cooperative control traffic network flow, where vehicle carbon emission and traffic efficiency are considered as reward functions to minimize carbon emissions and traffic congestions. In this method, we design a local coordination mechanism based on graph convolutional network to guide the multi-agent decision-making process by extracting spatial topology and traffic flow characteristics between adjacent intersections. This enables distributed agents to make low-carbon decisions which not only account for their own interactions with the environment but also consider local cooperation with neighboring agents. Further, we design a global coordination mechanism based on self-attention value decomposition to guide multi-agent learning process by assigning various weights to distributed agents with respect to their contribution degrees. This enables distributed agents to learn a globally optimal low-carbon control strategy in a cooperative and adaptive manner. In addition, we design a cloud computing-based parallel optimization algorithm for the GSAVD-MAC model to reduce calculation time costs. Simulation experiments based on real road networks have verified the advantages of the proposed method in terms of computational efficiency and control performance.