Article
Engineering, Civil
Weifeng Zhang, Zhe Wu, Xinfeng Zhang, Guoli Song, Yaowei Wang, Jie Chen
Summary: This paper proposes a novel traffic forecasting method that considers the traffic state trend and robust spatial relation, and utilizes temporal context information for local-period spatial relation analysis. It also introduces a temporal attention module to capture temporal features and a future feature inference module to infer future traffic information. Experimental results on real-world traffic datasets demonstrate that the proposed method outperforms other state-of-the-art methods.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Engineering, Multidisciplinary
Qin Zhang, Keping Yu, Zhiwei Guo, Sahil Garg, Joel J. P. C. Rodrigues, Mohammad Mehedi Hassan, Mohsen Guizani
Summary: This work proposes a graph neural network-driven traffic forecasting model for the connected Internet of vehicles (CIoVs), named Gra-TF. By utilizing ensemble learning and three typical graph-level prediction methods, an integrated and enhanced forecasting model is constructed to minimize uncertainty in CIoVs. Numerical results show that Gra-TF improves prediction accuracy by 30% to 40% compared to baseline methods.
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING
(2022)
Article
Energy & Fuels
Hakan Acikgoz
Summary: The study proposes a novel deep solar forecasting approach that combines multiple techniques and models to achieve high prediction accuracy. The experimental results show that the proposed method demonstrates accurate and robust forecasting performance, outperforming traditional regression models.
Article
Engineering, Civil
Xiaolei Ma, Houyue Zhong, Yi Li, Junyan Ma, Zhiyong Cui, Yinhai Wang
Summary: The study proposes a new framework for network-level traffic forecasting using CapsNet and NLSTM, demonstrating their superiority in capturing complex spatiotemporal traffic patterns through visualizing and quantitatively evaluating experimental results.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Huaying Li, Shumin Yang, Youyi Song, Yu Luo, Junchao Li, Teng Zhou
Summary: This study proposes a spatial dynamic graph convolutional network (SDGCN) for accurate traffic flow forecasting, addressing the challenges of complex traffic network spatial correlation and high nonlinear and dynamic traffic conditions. The SDGCN utilizes a dynamic graph generated by an attention fusion network to capture the changeable spatial correlation, and embeds dynamic graph diffusion convolution into gated recurrent unit to explore spatio-temporal dependency simultaneously. Experimental results on two public datasets demonstrate the superior performance of the network.
APPLIED INTELLIGENCE
(2023)
Article
Business, Finance
Mingxi Liu, Guowen Li, Jianping Li, Xiaoqian Zhu, Yinhong Yao
Summary: After constructing a feature system with 40 determinants affecting the price of Bitcoin, a deep learning method called SDAE was used for prediction. The SDAE model outperformed traditional methods such as BPNN and SVR in both directional and level prediction accuracy.
FINANCE RESEARCH LETTERS
(2021)
Article
Computer Science, Information Systems
Yang Pan, Xiao Zhang, Hui Jiang, Cong Li
Summary: This study proposes a classification method for network traffic based on graph convolution and Long-Short Term Memory (LSTM), which can effectively extract potential features and outperforms other methods in classification performance.
Article
Mathematical & Computational Biology
Keruo Jiang, Zhen Huang, Xinyan Zhou, Chudong Tong, Minjie Zhu, Heshan Wang
Summary: Multivariate time series (MTS) are important in daily life, and traditional forecasting methods for MTS are time-consuming and have limitations. This study proposes a deep learning network called DBI-BiLSTM, which outperforms traditional methods in prediction performance and feature extraction.
MATHEMATICAL BIOSCIENCES AND ENGINEERING
(2023)
Article
Engineering, Electrical & Electronic
Xuguang Hu, Dazhong Ma, Guanjun Meng, Rui Wang, Huaguang Zhang
Summary: This article proposes a cross-area knowledge learning (CKL) method for status monitoring in transportation-oriented energy interconnected system. The method uses a convolutional neural network to extract and separate features of different system status and improves the learning ability and classification accuracy for unlabeled samples through the sample classify module and virtual adversarial training item.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Automation & Control Systems
Chenhan Zhang, Shuyu Zhang, James J. Q. Yu, Shui Yu
Summary: This article proposes a novel federated learning framework that protects the topological information of transportation networks through a differential privacy-based adjacency matrix preserving approach, and introduces an adjacency matrix aggregation approach to allow local GNN-based models to access the global network for improved training effectiveness, while using an ASTGNN model for traffic speed forecasting.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2021)
Article
Chemistry, Analytical
Andrei-Cristian Rad, Camelia Lemnaru, Adrian Munteanu
Summary: Dot-product attention is a powerful mechanism for capturing contextual information. Efficient alternatives have been developed to overcome the bottleneck in constructing the attention map. This study compares these methods in the context of a spatio-temporal forecasting model, showing significant reductions in training and inference times while maintaining comparable performance.
Article
Computer Science, Artificial Intelligence
Shuai Zhang, Kun Zhu, Wenyu Zhang
Summary: A novel deep learning model is proposed to represent the spatial correlations of traffic network more effectively through a new correlation matrix structure. The model calculates the correlations among sensors to construct speed, volume, and occupancy correlation matrices, and optimizes the placement of highly correlated sensors using an enhanced heuristic optimization algorithm. The three optimal correlation matrices are then combined to form a three-dimensional multivariate correlation matrix characterized by locally high correlation, which enables the exploitation of deep spatial features of traffic network.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Computer Science, Information Systems
Javier Garcia-Siguenza, Faraon Llorens-Largo, Leandro Tortosa, Jose F. Vicent
Summary: In recent years, new Artificial Intelligence methods have been developed to improve the explainability and interpretability of models, particularly in dealing with black box machine learning methods. Our approach combines predictive and explainability techniques, showing that deep learning models have higher accuracy than statistical regression and classic machine learning models. Among various deep learning models, the Adaptive Graph Convolutional Recurrent Network performs the best in spatio-temporal traffic datasets. In terms of explainability, GraphMask achieves higher fidelity metric than other methods. Experimental results demonstrate that our approach enhances the accuracy of deep learning models, making them more transparent and interpretable, facilitating behavior analysis and improving model understanding.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Information Systems
Ruobin Gao, Wen Xin Cheng, P. N. Suganthan, Kum Fai Yuen
Summary: This paper utilizes machine learning algorithms and statistical methods to forecast the inpatient discharges of Singapore hospitals, and proposes an ensemble deep learning algorithm based on random vector functional links. Through optimizing the ensemble deep learning framework and validating it with multiple forecasting metrics and statistical tests, the superiority of the proposed algorithm is demonstrated.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2022)
Article
Engineering, Civil
Kadiyala Ramana, Gautam Srivastava, Madapuri Rudra Kumar, Thippa Reddy Gadekallu, Jerry Chun-Wei Lin, Mamoun Alazab, Celestine Iwendi
Summary: Traffic problems are worsening due to population growth in urban areas, causing issues such as air pollution, fuel consumption, law violations, noise pollution, accidents, and time loss. Traffic prediction is crucial in smart cities to reduce congestion, but current methods are not suitable for real-world applications. In this study, Vision Transformers (VTs) were used with Convolutional Neural Networks (CNN) to accurately predict traffic flow, particularly during abnormal situations. The proposed technology outperformed traditional methods in terms of precision, accuracy, and recall, while also promoting energy conservation through rerouting.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Automation & Control Systems
Yuxia Pan, Kaizhou Gao, Zhiwu Li, Naiqi Wu
Summary: This paper addresses a distributed lot-streaming permutation flow shop scheduling problem and proposes five meta-heuristics to solve it. Experimental results show that the artificial bee colony algorithm with improved strategies exhibits the best competitiveness for solving the problem with makespan criteria.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Telecommunications
Haijing Ning, Yisheng An, Yaxin Wei, Naiqi Wu, Chen Mu, Hanhan Cheng, Chenxing Zhu
Summary: This paper proposes a traffic warning message dissemination system (TWMDS) framework along with a protocol called the reverse routing protocol (RRP) for disseminating traffic warning messages (TWMs) in vehicular ad hoc networks (VANETs). TWMs are generated by vehicles causing abnormal traffic events and are only forwarded to potential congestion area (PCA) vehicles. Three colored petri net models are developed to analyze the interaction behavior of TWMDS. Simulation results using the Veins framework show the higher application value of RRP compared to other routing protocols in TWMDS. TWMDS enables drivers to quickly re-route their paths to alleviate traffic congestion.
VEHICULAR COMMUNICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Yisheng An, Yuxin Gao, Naiqi Wu, Jiawei Zhu, Hongzhang Li, Jinhui Yang
Summary: As the number of electric vehicles increases, scheduling the charging operations of EVs in urban areas becomes an important issue. This paper investigates the EV charging problem at the scheduling level and proposes a mathematical model and scheduling algorithms to improve charging efficiency.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Automation & Control Systems
HuiChu Fu, Yan Qiao, LiPing Bai, NaiQi Wu, Bin Liu, YunFang He
Summary: Semiconductor manufacturing relies on a complex production line with over 1,000 tools and an overhead hoist system. IoT and software engineering technologies are adopted to collect data from tools and ensure cooperation between tools and the hoist system. However, the current fault detection and classification (FDC) system has limitations in data-processing capacity, leading to delayed detection of undesired events. This study proposes a new FDC framework based on the Hadoop ecosystem to overcome these limitations and improve diagnosis efficiency, while presenting a migration path for a smooth transition without shutting down the wafer fabrication line. Experimental results demonstrate the safe and stable operation of the proposed FDC framework.
IEEE ROBOTICS & AUTOMATION MAGAZINE
(2023)
Article
Automation & Control Systems
Yuxia Pan, Kaizhou Gao, Zhiwu Li, Naiqi Wu
Summary: A distributed flow-shop scheduling problem with lot-streaming is addressed in this paper. A biobjective mathematic model is developed and an improved Jaya algorithm is proposed to solve the problem. Experimental results show that the strategies designed for the algorithm are competitive for solving the problem with makespan and total energy consumption criteria.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Automation & Control Systems
WenQing Xiong, Jie Li, Yan Qiao, LiPing Bai, BaoYing Huang, NaiQi Wu
Summary: Nowadays, cluster tools are extensively used in wafer manufacturing processes. With the development of equipment design, multifunctional process modules (MPMs) are equipped to serve for processing multiple operations together. An efficient scheduling method is desired to quickly adapt to wafer processing parameter changes and maximize productivity. A deadlock-free Petri net (PN) model is developed and two algorithms are proposed to calculate the makespan. An adaptive scheduling method is presented to minimize the makespan by setting the functions of MPMs. Experimental results show the efficiency and effectiveness of the proposed method.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Computer Science, Information Systems
Tairan Song, Yan Qiao, Yunfang He, Naiqi Wu, Zhiwu Li, Bin Liu
Summary: Cluster tools are important equipment in semiconductor manufacturing systems and widely used for wafer fabrication processes. The scheduling of cluster tools with reentrant processes is complex. Existing studies only provide optimal 1-WP schedules for dual-arm cluster tools with two-time reentering. This work explores the existence of 1-WP schedules for dual-arm cluster tools with more than two reentering times and proposes new methods for three-wafer periodical schedules.
Article
Mathematics
Yuanxiu Teng, Zhiwu Li, Li Yin, Naiqi Wu
Summary: Differential privacy is introduced to discrete event systems modeled by probabilistic automata to protect state information pertaining to system resource configurations. State differential privacy is defined to protect the initial state of a discrete event system. A step-based state differential privacy verification method is proposed to make it difficult for an attacker to determine the initial state from which a system evolves. Experimental studies show that the proposed method can effectively verify state differential privacy and enforce privacy protection.
Article
Engineering, Electrical & Electronic
Yan Qiao, Naiqi Wu, Zhiwu Li, Abdulrahman M. Al-Ahmari, Abdul-Aziz El-Tamimi, Husam Kaid
Summary: This work addresses the scheduling problem of automotive glass manufacturing systems and proposes an efficient solution method. By determining the minimal size of each batch and using integer linear programming and a polynomial algorithm, the computational complexity is greatly reduced.
Article
Computer Science, Cybernetics
Shaohua Teng, Chengzhen Ning, Wei Zhang, NaiQi Wu, Ying Zeng
Summary: This article proposes a fast asymmetric and discrete cross-modal hashing (FADCH) method to address the issues in supervised cross-modal retrieval. It leverages matrix factorization to construct a common semantic subspace, aligns it with semantic representation, embeds labels into hash codes, and uses an asymmetric strategy with relaxation to associate hash codes with semantic representation. Experimental results on benchmark datasets demonstrate the superiority of the FADCH method.
IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS
(2023)
Article
Automation & Control Systems
QingHua Zhu, GengHong Wang, NaiQi Wu, Yan Qiao, Yan Hou, MengChu Zhou, SiDe Zhao
Summary: In this study, a scheduling method is developed for the concurrent fabrication processes of two different wafer types using a multicluster tool, considering wafer residency time constraints. The proposed approach converts a one-wafer cyclic schedule into a one-wafer-per-type cyclic schedule using a backward strategy based on a single wafer type and reveals its temporal properties. Several necessary and sufficient conditions are derived for smooth operation of a single-arm multicluster tool system and synchronization of multiple robots. Two efficient algorithms are proposed to determine the feasibility of a periodic schedule and obtain a schedule that achieves the lower-bound cycle time under a two-backward strategy, maximizing the productivity of the multicluster tool. Numerical simulations and practical examples are presented to demonstrate the applications and performance of the proposed approach.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Computer Science, Information Systems
Tairan Song, Yan Qiao, Yunfang He, Jie Li, Naiqi Wu, Bin Liu
Summary: In modern semiconductor manufacturing, the equipment automation program (EAP) is a crucial system that needs to be enhanced for better stability and compatibility. This study presents a new framework for a distributed EAP system with new technologies, aiming to solve the problems of traditional EAP and make it more adaptable and scalable.
Article
Engineering, Civil
Zhongjie Lin, Kaizhou Gao, Naiqi Wu, Ponnuthurai Nagaratnam Suganthan
Summary: This paper proposes a novel hybrid algorithm framework that combines meta-heuristics with Q-learning to solve the urban traffic light scheduling problems (UTLSP) with eight phases for the first time. The framework includes a mathematical model to describe UTLSP, five improved meta-heuristics, five local search operators, and two Q-learning-based ensemble strategies. Experimental results validate the effectiveness of the proposed ensemble strategies and show that the improved water cycle algorithm with the first Q-learning strategy performs the best in solving the considered problems.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Engineering, Civil
Yaxin Wei, Haijing Ning, Yisheng An, Naiqi Wu, Xiangmo Zhao
Summary: This paper addresses the issue of designing safety Petri net-based controllers to prevent vehicle flow deadlocks at intersections. The designed controller can monitor and guide the flow of vehicles at an intersection to pass safely without causing deadlocks. The study investigates intersection deadlock scenarios, analyzes the physical size of right-of-way cells, develops an initial Petri net model, and proposes a deadlock prevention strategy for controller design. The effectiveness of the strategy is illustrated through examples and theoretical proof. The study contributes to the advancement of safety controllers for self-driving vehicles at intersections.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Yuanxiu Teng, Li Yin, Zhiwu Li, Naiqi Wu
Summary: This research introduces differential privacy to protect the initial state of probabilistic discrete event systems. A state differential privacy verification method is proposed to determine the similarity of probability distributions between adjacent initial states. A supervisory control method is also proposed for systems that do not satisfy state differential privacy.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)