Article
Computer Science, Artificial Intelligence
Kun Tang, Tangyi Guo, Fei Shao, Yongfeng Ma, Aemal J. Khattak
Summary: Urban road networks are increasingly congested, which affects both transport efficiency and the living environment. To address these issues, it is crucial to provide proactive knowledge of traffic performance. However, challenges exist due to the uncertainties of signalized urban road networks, such as limited approaches and the use of traffic flow parameters as proxies for traffic state. This study developed a dynamic factor model-based approach to forecast multi-step network traffic states collaboratively and addressed these gaps by considering co-movement dynamics and region-specific distinctions.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Zhuhua Liao, Haokai Huang, Yijiang Zhao, Yizhi Liu, Guoqiang Zhang
Summary: Urban planning and function layout have significant implications for commuter journeys. This paper proposes a new method that utilizes ride-hailing trajectories to analyze and forecast traffic flow among functional areas. The method extracts spatial correlations through graph convolutional neural networks and temporal features through an attention-based gated graph convolutional network.
ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION
(2023)
Article
Robotics
Tong Qin, Haihui Huang, Ziqiang Wang, Tongqing Chen, Wenchao Ding
Summary: An accurate road topological structure is crucial for autonomous driving in complex urban environments. Currently, a High-Definition map (HD map) is heavily relied upon by most autonomous vehicles to navigate through the city. However, producing such a map manually is time-consuming and prone to errors. In this letter, a framework is proposed to automatically generate the topological map of complicated intersections using crowd-sourced semantic information about the environment and traffic flows. This highly automatic and scalable framework speeds up HD map production and reduces costs, with results comparable to traditional HD maps validated by real-world crowdsourcing data.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Computer Science, Interdisciplinary Applications
Yuhong Gao, Zhaowei Qu, Xianmin Song, Zhenyu Yun
Summary: The determination of traffic carrying capacity and the proposal of a novel calculation model for carrying capacity are the main objectives of this study. It is found that the service level of road network is the key factor affecting the number of vehicles. Based on the definition of carrying capacity and the relationship between average travel speed and service level, a calculation model is established and verified using VISSIM simulation software, which provides accurate results.
SIMULATION MODELLING PRACTICE AND THEORY
(2022)
Article
Automation & Control Systems
Teresa Pamula, Renata Zochowska
Summary: In this article, a new method for predicting OD matrix based on traffic data using deep learning is proposed. The method eliminates the need for complex data acquisition and processing, and achieves high accuracy and resistance to missing data. A case study conducted in a medium-sized city in Poland demonstrates the practical implementation potential in real-time traffic assignment systems. The method does not require questionnaire research or detailed spatial development information.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Mathematics
Kai Zhang, Zixuan Chu, Jiping Xing, Honggang Zhang, Qixiu Cheng
Summary: This study proposes a data-driven model for predicting traffic congestion by analyzing the spatio-temporal features of traffic flow. The model utilizes the traffic zone/grid method to store the average speed of vehicles on local area roads and employs a discrete snapshot set to characterize the spatial and temporal features of traffic flow. By transforming the global urban transportation network into traffic zones, the evolution of traffic congested flow in various time dimensions is examined. The model incorporates a convolutional LSTM network to learn the temporal and local spatial features of traffic flow, while a convolutional neural network effectively captures the global spatial features. Numerical experiments on two cities' transportation networks demonstrate that the proposed model outperforms traditional traffic flow prediction models.
Article
Engineering, Civil
James J. Q. Yu, Christos Markos, Shiyao Zhang
Summary: The proposed model incorporates graph deep learning techniques to predict spatial-temporal data correlation of traffic dynamics, demonstrating consistent improvements over existing methods in case studies.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Engineering, Civil
Yang Xu, Duo Jia Zhang, Xin Zhang, Kin Keung Lai, Bing Su
Summary: This study established a catastrophe model of the urban road system under occasional congestion using cusp catastrophe theory, and discussed the feasibility of congestion control based on the catastrophe characteristics. It was found that the control method based on catastrophe characteristics could effectively improve the efficiency of the road system theoretically.
JOURNAL OF ADVANCED TRANSPORTATION
(2021)
Article
Green & Sustainable Science & Technology
Song Fang, Linghong Shen, Jianxiao Ma, Chubo Xu
Summary: This paper discusses the setting method of variable lane boundaries in urban tunnels to alleviate the influence of low-speed vehicles on tunnel safety. The research results show that allowing lane changes can significantly alleviate the impact of low-speed vehicles on tunnel traffic flow.
Article
Computer Science, Artificial Intelligence
Matheus A. C. Alves, Robson L. F. Cordeiro
Summary: Recent works in traffic forecast rely on burdensome procedures and various additional data, while our proposed AdaptFlow algorithm can accurately predict future traffic flow by monitoring locally connected highways only.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Information Systems
Peixiao Wang, Yan Zhang, Tao Hu, Tong Zhang
Summary: This study proposes a dynamic temporal graph neural network model that considers missing values and dynamic spatial relationships for urban traffic flow prediction. The model achieves good prediction results and outperforms existing baselines on a real traffic dataset.
INTERNATIONAL JOURNAL OF GEOGRAPHICAL INFORMATION SCIENCE
(2023)
Article
Computer Science, Interdisciplinary Applications
Zhen Di, Lingxuan Li, Mengfei Li, Shenghu Zhang, Yuxiao Yan, Mengfei Wang, Bin Li
Summary: Most carbon emissions in cities come from human activities related to the urban economy, construction, and transportation. Underground logistics systems and coordinated operation modes involving metro-based freight can significantly reduce carbon emissions in urban transportation, as demonstrated by the case study of Beijing Metro Batong Line.
COMPUTERS & INDUSTRIAL ENGINEERING
(2023)
Article
Chemistry, Multidisciplinary
Xing Xu, Hao Mao, Yun Zhao, Xiaoshu Lu
Summary: In this paper, a causal gated low-pass graph convolutional neural network (CGLGCN) is proposed for traffic flow prediction. The network combines a self-designed low-pass filter to extract spatial features more comprehensively and improve prediction accuracy.
APPLIED SCIENCES-BASEL
(2022)
Article
Green & Sustainable Science & Technology
Amin Mallek, Daniel Klosa, Christof Bueskens
Summary: This study examines the impact of data loss on the behavior of the k-nearest neighbors model used for traffic flow forecasts. Three efficient techniques for reconstructing missing data are proposed, and the performance of the model is evaluated under different datasets. The results show that the model can accurately predict traffic flow with a low error rate.
Article
Computer Science, Artificial Intelligence
Dung David Chuwang, Weiya Chen, Ming Zhong
Summary: Global urbanization has increased the importance of urban rail transit systems, and this study proposes a fusion strategy using machine learning methods to improve prediction accuracy by considering the complex pattern of passenger flow.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Information Systems
Fengjun Xiao, Mingming Lu, Ying Zhao, Soumia Menasria, Dan Meng, Shangsheng Xie, Juncai Li, Chengzhi Li
HUMAN-CENTRIC COMPUTING AND INFORMATION SCIENCES
(2018)
Article
Computer Science, Information Systems
Soumia Menasria, Jianxin Wang, Mingming Lu
WORLD WIDE WEB-INTERNET AND WEB INFORMATION SYSTEMS
(2018)
Article
Engineering, Electrical & Electronic
Kongyang Chen, Guang Tan, Jiannong Cao, Mingming Lu, Xiaopeng Fan
IEEE SENSORS JOURNAL
(2020)
Article
Materials Science, Multidisciplinary
Yunqiang Wang, Mingming Lu, Zi Wang, Jin Liu, Lei Xu, Zijun Qin, Zexin Wang, Bingfeng Wang, Feng Liu, Jianxin Wang
Summary: High-throughput experiments on superalloys have become common, generating large amounts of data. A machine learning model was designed to automate the experimental analysis process, using the Unet algorithm to segment precipitated phases from superalloy images and a regression algorithm to predict morphological parameters based on their composition. This method may offer guidance for future superalloy composition design.
MATERIALS & DESIGN
(2021)
Article
Environmental Sciences
Mingming Lu, Qi Li, Li Chen, Haifeng Li
Summary: Patch-Noobj is an adversarial attack method that generates universal adversarial patches based on the size of attacked aircraft in remote sensing images. The experiment shows that these patches effectively reduce the Average Precision of the YOLOv3 detector on various datasets, demonstrating their attack transferability.
Article
Computer Science, Hardware & Architecture
Mingming Lu, Zhixiang Xiao, Haifeng Li, Ya Zhang, Neal N. Xiong
Summary: The paper presents a Feature Pyramid-based Graph Convolutional Neural network for Graph Classification (FPGCN-GC) to reduce information loss through multi-scale hierarchical information fusion and achieve adaptive feature fusion with a learnable weighted residual connection and self-attention mechanism.
JOURNAL OF SYSTEMS ARCHITECTURE
(2022)
Article
Environmental Sciences
Mingming Lu, Yongchuan Xu, Haifeng Li
Summary: This paper discusses the challenges brought by high-resolution remote sensing images to traditional vision tasks, and the application and advantages of vehicle re-identification in remote sensing images. To address the existing issues, the authors construct a large-scale vehicle re-identification dataset based on UAV views and propose a Global Attention and full-Scale Network (GASNet) model. Experimental results show that GASNet outperforms baseline models in terms of Rank-1 and mAP on the dataset.
Article
Engineering, Civil
Aikun Xu, Ping Zhong, Yilin Kang, Jiongqiang Duan, Anning Wang, Mingming Lu, Chuan Shi
Summary: This paper presents a multi-modal transportation recommendation algorithm based on a carefully constructed Heterogeneous graph Attention Network (THAN). The algorithm utilizes a novel graph embedding method, constructs a heterogeneous graph from large-scale data, and employs a hierarchical attention mechanism and a fusion neural layer for node embedding and transport mode prediction.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Information Systems
M. Lu, Y. Wang, D. Tan, L. Zhao
Summary: Source code mining focuses more on project codes, which are usually large and standardized, than student programs, which are smaller and more irregular. Therefore, specific methods are needed to improve the efficiency of information extraction in student programs.