Search

In the process of collecting traffic information, traditional traffic flow control methods have some problems, such as high loss rate of sensing and detection information and long average queue length, which lead to the unsatisfactory effect of traffic control and evacuation in main roads. Therefore, based on multi-sensor information fusion technology, this study designs a new intelligent control method of traffic flow on main road. According to the multi-sensor information, the fusion algorithm and measurement equation are designed to obtain the main road traffic sensor position, vehicle linear velocity and other parameters, and update the fusion results in real time. Then, the short-term characteristics of traffic flow are used to control the appearance time of target section. Based on the ordinary differential equation, the main road traffic network is represented by intelligent directed graph, and the intelligent dredging model of urban main road traffic flow is constructed. The simulation results show that in five experimental roads, the average delay time of this method is less than 35 s, the maximum average number of stops is only 8.8, the average queue length is less than 30 m, the average travel speed is more than 25 km/h, the traffic flow per unit time is more than 250veh/10 min, and the traffic congestion index of main road is always more than 5.5. The above indexes of this method are better than those of traditional methods, which verifies that this method has better application performance. [ABSTRACT FROM AUTHOR]

Autonomous mining transportation is an intelligent traffic control system that can provide better economics than traditional transportation systems. The velocity trajectory of a manned vehicle depends on the driver's driving style. Still, it can be optimized utilizing mathematical methods under autonomous driving conditions. This paper takes fuel and electric mining vehicles with a load capacity of 50 tons as the subject. It contributes a multi-objective optimization approach considering time, energy consumption, and battery lifetime. The dynamic programming (DP) algorithm is used to solve the optimal velocity trajectory with different optimization objectives under two types of mining condition simulation. The trajectories optimized by the single objective, energy consumption, usually adopt the pulse-and-gliding (PnG) approach frequently, which causes the battery capacity loss and increases the travel time. Hence, a multi-objective optimization approach is proposed. For electric vehicles, trajectories optimized by the multi-objective approach can decrease the battery capacity loss by 22.01 % and the time consumption by 41.28 %, leading to a 42.12 % increment in energy consumption. For fuel vehicles, it can decrease the time consumption by 32.54 %, leading to a 7.68 % increment in energy consumption. This velocity trajectory is smoother with less fluctuation. It can better meet the requirements of mining transportation and has a particular reference value for optimizing autonomous transportation costs in closed areas. [ABSTRACT FROM AUTHOR]

Accurately predicting network traffic is helpful for improving a variety of spatial-temporal data mining applications, such as intelligent traffic control, network planning and anomaly detection. The mainstream graph-based methods are limited by the node-level message passing mechanism and require transforming dimensions to generate edge representations. Accordingly, some researchers propose using edge convolution to directly learn edge representations for network traffic prediction. However, node-level and edge-level information aggregation are two different perspectives, and their combination of them can achieve better performance. This paper proposes a novel model for network traffic prediction named the Enhanced Edge Convolution-based Spatial-Temporal Network (EESTN). Armed with a Graph Neural Network and Hypergraph Neural Network, EESTN employs the edge convolutions defined on the graph and hypergraph to effectively extract spatial features. EESTN further combines node convolution to capture the complex correlations among nodes and utilizes an attention mechanism to generate the edge convolution kernel for the decoder. Moreover, a 3D convolution-based multihead self-attention mechanism and a hierarchical loss function are proposed to capture the long-term temporal dependence and make full use of the model's represent ability. Finally, we conduct extensive experiments to validate the effectiveness of EESTN, and the related results demonstrate that EESTN outperforms the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

In recent years, the sharp increase in the number of residential motor vehicles and non-motorized vehicles has led to a huge challenge in the intelligent control of urban road traffic operation status. To address the problems of low accuracy of small target detection in complex traffic scenes and the difficulty of balancing the accuracy and detection speed of multiple targets, we propose a deep learning-based Semantic interpolation target detection network SI-SSD based on feature pyramid network FPN and feature graph fusion method, and transfer it to target detection in complex traffic scenes. Shallow feature maps have multi-edge information and deep feature maps have high semantic information. With this idea in mind, the high-resolution shallow feature layer Conv4_3 for the detection of small targets and the deep feature layer Conv7_2 with rich semantic information in SSD network are fused, so as to enhance the sensitivity of the network to the feature information of small targets and improve the network's ability to detect small targets in complex traffic scenarios. The experimental results show that the improved SI-SSD network model has good robustness in the detection of relatively small targets, which can ensure the detection speed and improve the accuracy of detection and recognition significantly. Tests of the data sets UA-DETRAC and VOC2007 show that in the detection of relatively small targets such as Motorbike and Pedestrian, the accuracy is 89.3% and 86.7%, respectively, and the mean average precesion (mAP) is 89.05%. Compared with other target detection networks based on deep learning, the average accuracy of all types of target detection is improved by 4%, and the detection speed is maintained at about 32 fps. [ABSTRACT FROM AUTHOR]

Accurate vehicle speed prediction is of great significance to the urban traffic intelligent control system. However, in terms of traffic speed prediction, the modules that integrate temporal and spatial features in the existing traffic speed prediction methods are effective in short-term prediction, but the medium-term or long-term prediction errors are relatively large. In order to reduce the errors of existing methods in short-term prediction and predict the medium-term and long-term traffic speed, this paper proposes a traffic speed prediction method that combines attention and Spatial–temporal features, referred to as ASTCN. Specifically, unlike previous methods, ASTCN can use the temporal attention convolutional network (ATCN) to separately extract temporal features from the traffic speed features collected by each sensor, and use the spatial attention mechanism to extract spatial features and then perform spatial–temporal feature fusion. Experiments on three real-world datasets show that the proposed ASTCN model outperforms the state-of-the-art baselines. [ABSTRACT FROM AUTHOR]

This paper introduces a design and analysis method of the intelligent multiple-vehicle system (IMVS) based on preview control in the environment of the vehicle to vehicle (V2V). Considering the time delay, the vehicle longitudinal dynamics system is constructed. The information flow between vehicles in the platoon is realized by the onboard sensors and V2V. Based on the actual data sampling and processing, the cooperative optimal preview control problem of the vehicle platoon stability system is discretized. The error is used as a term of the state vector to expand, and the original problem is transformed into an augmented global optimal control problem. The existence and stability of the controller solution are analyzed. The asymptotically stable optimal control law is obtained by using the state augmented matrix, and the global optimal preview controller of original platoon stability is obtained. The simulation results verify the effectiveness of the designed controller. The following error decreases with the increase of the preview step, which ensures the stability of the platoon. From the first follower to the tenth one, the average spacing error decreases by 92.9 %, and the average following error of velocity decreases from 3.97 to 1.56 %. [ABSTRACT FROM AUTHOR]

The Software Defined Networking (SDN) promises exciting new networking functionality. However, there is always remains a chance of programming errors that result in unreliable data communication. The centralized programming model helps decrease bugs' probability where a single controller manages the whole network. Yet, many real-time events occur at switches and end hosts, which often affect and add a delay in the communication process. One of those events includes unannounced destination host migration after installing flow rules during receiving of data packets. Such destination host movement results in the loss of packets because controller is not aware of this recent event. Therefore, we need an efficient approach to transmit packets without any packet loss despite destination host migration. This paper proposed a design to achieve the objective as mentioned above by defining a layer named Intelligent Transmission Control Layer (ITCL). It monitors all the end hosts' connections at their specific locations and performs necessary actions whenever the connection state changes for one or multiple hosts. The controller collects information of end nodes and state change through ITCL using A star search algorithm. After that, it updates flow tables accordingly to accommodate a location-change scenario with a route-change policy. ICTL is developed on prototype-based implementation using a popular POX controller platform. By comparing ITCL with the existing solution, we conclude that our proposed approach exhibits efficient performance in terms of Packet loss, Bandwidth usage, and Network Throughput. [ABSTRACT FROM AUTHOR]

Software-defined networking (SDN) has become one of the most promising paradigms to manage large scale networks. Distributing the SDN control proved its performance in terms of resiliency and scalability. However, the choice of the number of controllers to use remains problematic. A large number of controllers may be oversized inducing an overhead in the investment cost and the synchronization cost in terms of delay and traffic load. However, a small number of controllers may be insufficient to achieve the objective of the distributed approach. So, the number of used controllers should be tuned in function of the traffic charge and application requirements. In this paper, we present an intelligent and resizable control plane for software defined vehicular network architecture, where SDN capabilities coupled with deep reinforcement learning (DRL) allow achieving better QoS for vehicular applications. Interacting with SDVN, DRL agent decides the optimal number of distributed controllers to deploy according to the network environment (number of vehicles, load, speed etc.). To the best of our knowledge, this is the first work that adjusts the number of controllers by learning from the vehicular environment dynamicity. Experimental results proved that our proposed system outperforms static distributed SDVN architecture in terms of end-to-end delay and packet loss. [ABSTRACT FROM AUTHOR]