ABSTRACT. As blockchain technology continues to be widely adopted, system scalability and performance issues have become increasingly prominent. To address the growing demand for transactions and enhance transaction processing efficiency, Layer-2 scaling technologies have become a critical research focus. Among these, Rollup technology is considered one of the most effective solutions to the scalability challenges of blockchains. Rollups process transactions off-chain and submit the results on- chain, thereby significantly freeing up on-chain storage space and greatly improving blockchain throughput. However, mainstream Rollup systems currently rely on centralized sequencer nodes to manage transaction ordering and batching. While these centralized sequencer nodes simplify system design and implementation, they introduce several security and reliability concerns. Firstly, centralized sequencer nodes are susceptible to censorship attacks, which pose risks of transaction tampering or denial for the system and its users. Secondly, any failure of centralized sequencer nodes can paralyze the entire network, severely affecting system availability and stability.To address these issues, we propose an innovative dynamic weighted random leader node election algorithm. This algorithm incorporates a commitment phase, a multi-signature voting phase, and a dynamic election pool management mechanism to ensure the security, privacy, and fairness of the election process. Additionally, by integrating elements of randomness, this algorithm increases the unpredictability of the election process, thereby enhancing the system's resilience to attacks. Experimental results demonstrate that the our algorithm significantly outperforms existing methods in filtering out abnormal nodes, thereby improving the reliability and security of sequencer node election.

ABSTRACT. Estimating the Worst-Case Execution Time (WCET) of programs in an embedded multi-core environment is fundamental for schedulability analysis. In this paper, we propose a framework for calculating the WCET of programs originally designed for single-core systems when migrated to and executed in parallel on multi-core environments. First, we decompose single-core tasks and convert them into parallel programs using OpenMP. Our approach extends the task dependency graph to describe relevant information for parallel tasks (such as execution time estimates, core allocation, and inter-task dependencies). Finally, we present an iterative algorithm based on abstract interpretation. This algorithm takes the parallel task dependency graph as input and uses fixed-point analysis to determine the WCET of parallel programs. We conducted experiments on a physical machine to verify the feasibility of our solution.

ABSTRACT. The irregular behavior of ship pilots on duty is a potential unsafe factor for ship navigation.Identifying abnormal behaviors of ship pilots, reducing water traffic accidents caused by human factors, and improving the safety supervision capability of vessel navigation have become important problems in water traffic safety. In order to be closer to the real scene, a private dataset is established with a full-field-of-view simulated bridge as the background, divided into 6 actions, totaling 1560 videos; RGB, IR, and Depth modalities are used as inputs to compensate for the problems of single modality being easily affected by lighting and lacking three-dimensional information; a lightweight modified 3DCNN is designed as the main feature extraction network to extract modal features; Double Attention is utilized for multimodal fusion, enhancing features by overlaying the same features from different modalities, thus better leveraging the complementarity and correlation between different modalities. The research results show that the developed model has good performance and can effectively identify different behaviors, achieving an overall accuracy of 92.95% on the custom dataset, with the behaviors of texting with a phone and using a telescope reaching recognition rates of 98% and 94%, respectively, and an F1-Score of 0.8731.

ABSTRACT. This study successfully applies the spatiotemporal transformer-based model to sea surface height anomaly prediction task. An improved ResNet autoencoder, named SW-GD Autoencoder, is proposed based on Sliced-Wasserstein distance and Gradient Difference loss functions to extract spatial feature information. This autoencoder demonstrates a reduced complexity and training difficulty compared to GAN-based architectures while maintaining high data generation quality. Additionally, the VidHRFormer model is employed for inference, introducing a large receptive field spatial feed-forward network module that enhances information transmission between adjacent patches in the feature map, thereby improving the observability of prediction results. To simplify training, we used a two-stage approach: first training the autoencoder separately, then combining it with VidHRFormer for final training. Compared to state-of-the-art ConvLSTM models, our method yields superior results.

ABSTRACT. Although moving block mode “hard-wall” safety protection principle adopted by the existing urban rail train control system has shortened the following interval of trains to a certain extent, it fails to meet the operational demand during peak hours. In this paper, a train following safety protection method based on stochastic reachable set is proposed, which upgrades the safety protection criterion from position non-overlapping in spatial dimension to trajectory non-overlapping in space-time dimension. Firstly, the whole process predicted trajectory of the preceding train and the safety braking trajectory of the following train are represented. Secondly, the efficient calculation model of collision probability is established by combining the probability distribution functions of the preceding train and the following train. Then, the corresponding safety protection curve is generated according to the safety verification results of the urban rail train. Finally, the method is verified by using the data of Beijing Metro Line 11. The experimental results show that, compared with the short-time protection, the method proposed in this paper can extend the reverse calculation starting point of the safety protection curve for following train and shorten the following interval on the basis of collision avoidance, which proves the necessity and applicability of this method.

ABSTRACT. Surface defect detection technology is a vital component of the steel industry that has garnered significant attention from the academic community in recent times. While modern methods with deep learning-based object detection provide better detection accuracy than traditional approaches, production requirements are often not met by their inference time. This study suggested an improved method based on the industry leading object detection algorithm YOLOv8 to address this issue and effectively detect steel surface defects while keeping a reasonable balance between speed and accuracy. First, a Stem module based on YOLOv8s is added to downsample the image and lessen the effect of redundant features in the original image. In the meantime, the backbone network gains a large kernel depthwise convolution module called InceptionNeXt+, which improves the model's ability to extract features and has a larger effective receptive field. Lastly, the base model's SPPF module is changed to a RsimSPPF module, which increases the model's accuracy and speeds up inference. The improved YOLOv8 model confirms the efficacy of the improved method with experimental results on the NEU-DET dataset achieving a mAP of 76.9%, increasing by 3.3 compared to the baseline model.

ABSTRACT. The problem of output constraints for a flexible wing system is discussed in this paper, taking into account that all constraint references come from an exosystem with unknown initial values. A partial differential equation-ordinary differential equation (PDE-ODE) coupled observer is developed for exponential estimation of the uncertain constraints. An observer-based nonlinear control method is proposed to restrain the bending and twisting displacements. Numerical simulation is carried out to demonstrate the effectiveness of the designed controls.

ABSTRACT. Currently, factors such as hidden health hazards, unhealthy lifestyles, and environmental pollution have led to the frequent occurrence of diseases, causing a common problem - a sharp increase in hospital admissions. However, for emergency departments (EDs), their medical resources and scale are insufficient to meet the challenges of this conventional hospital level. Thus, modeling the diagnostic process of the ED and applying appropriate method to predict the probability of discharges within a certain time is of significant importance for achieving rational resource allocation. Effective allocation of medical resources can ensure that patients receive timely and effective medical care, improve the quality and level of medical services, and thereby enhance patient satisfaction with medical care. Therefore, this paper chooses to use stochastic timed Petri nets (STPN) to model the workflow of the ED, in order to more accurately simulate various random events and delays that occur within the ED. Additionally, for better analysis of the model, we will convert the time interval values assigned to transitions into time interval values assigned to places, creating a Petri net known as TSdPN. Using Net learning method to predict patient discharge probabilities helps healthcare providers allocate beds, resources, and follow-up medical services more accurately, thereby improving the efficiency of medical resource utilization. Combining these two approaches makes ED operations more intelligent and efficient, providing patients with a better quality medical experience.

ABSTRACT. Pantograph is the crucial component of rail transit vehicles. Detecting and resolving faults swiftly is important in the pantograph. Existing fault diagnosis technologies with artificial intelligence have difficulties in dealing with some rare faults due to the scarcity of training samples. This paper proposes a new algorithm for few-shot fault diagnosis of pantograph. The proposed method constructs a reconstructive adversarial network with hierarchical structure to mine multi-scale feature from few-shot samples. A hierarchical data generation module is designed leveraging adversarial learning to generate scarce category samples. An initial classifier is employed to constrain the generated results, while the classifier is dynamically reconstructed to enhance overall performance. This paper designs detailed experiments to demonstrate that, compared with other existing methods, the proposed algorithm can better solve the problem of pantograph fault diagnosis.

ABSTRACT. In this paper, presents a lightweight human pose estimation algorithm, M-AlphaPose, based on AlphaPose. Firstly, we integrate YOLOV5s for human detection, and use MobileNetV3-S as the backbone for pose estimation. Finally,achieving an effective system for estimating full-body poses of multiple individuals. The study conducts ablation experiments to analyze the impact of different components and compares the performance with other algorithms, emphasizing the importance of lightweight design. Compared to traditional AlphaPose, MAlphaPose significantly reduces computational complexity and parameter count while maintaining high detection accuracy, achieving an AP50 score of 0.921.

ABSTRACT. Occupational pneumoconiosis(OP) staging is a standard-oriented image classification process that relies on patients' chest X-rays. However, the pathogenesis of OP affects its staging accuracy from two aspects: a) flocculent pneumoconiosis lesions are difficult to be extracted from X-rays, and b) the stages of OP are often extremely imbalanced, significantly diminishing the effectiveness of classification models that assume a balanced data distribution. These factors lead to inaccurate staging results. and significantly reduce the clinical value of existing models. To achieve accurate OP staging, we propose a Prior Knowledge-embedded OP staging framework which can enhance OP lesions with a dark channel prior-inspired method and handle imbalanced data with a triple-branch model. Empirical studies on ethically reviewed datasets indicate that the proposed framework performs more accurately in OP staging than state-of-the-art methods.

ABSTRACT. Neuromorphic electronic devices have already exhibited tremendous potential in mimicking the basic sensory and neuro-memory functions of living beings. However, while olfactory sensing electronics with ideal sensitivity and high selectivity achieved significant breakthroughs, synaptic devices that mimic the olfactory system have rarely been reported. Herein, a stretchable resistive olfactory neuromorphic device is proposed, which combines the common ductility of stretchable gold nanowires and porous polymer films to enable the olfactory synaptic function under different stretching states. The device achieves NO2-mediated synaptic plasticity and realize typical synaptic behaviors, including functions such as paired-pulse facilitation (PPF), short-term potential (STD) and long-depression (LTD), and is able to mimic the cumulative damage that can be caused by exposure to NO2 even in the tensile strains up to 50%. The proposed stretchable electrons pave a new way for next-generation olfactory biomimetic systems.

ABSTRACT. The tamper-proof feature of blockchain makes the correctness of smart contracts crucial before they are run on the chain. Using model checking methods to verify smart contracts helps ensure their correctness. Methods for converting smart contracts into semantically rigorous formal models that can be used in model checking methods still face the challenges of poor usability and difficulty in breaking away from manual modeling. Colored Petri nets, as discrete state-based systems, have similarities to smart contracts and are suitable for analyzing interactions between transactions. The analysis of control flow and data flow interactions, and transactions on the blockchain is missing in the research on converting smart contracts into colored Petri nets. In this paper, we propose an automated modeling method based on colored Petri nets, using template modeling and combination with data flow interactions and transaction modeling. Firstly, the data and control flows are modeled and combined, and secondly, the external user invocation is modeled to be able to correctly portray the on-chain contract invocations. The correctness in the smart contract is then verified based on the colored Petri net. Finally a prototype tool was implemented to demonstrate the usability and correctness of the modeling algorithm.

ABSTRACT. With the continuous advancements in science, technology, and production methods, the digitalization process of the chemical industry is rapidly progressing. This paper addresses the limitations of existing hierarchical system structures in chemical process virtual factory construction, which often result in lengthy development cycles and limited scalability, failing to accommodate real-time user needs. Additionally, conventional virtual factories primarily presented through computer screens lack immersion and interactivity. To overcome these challenges, this study proposes a de-layered and de-centralized virtual factory construction scheme. It integrates projected virtual reality technology to enhance personnel training and work experiences. This framework empowers users to develop customizable functional modules, thereby catering to personalized training requirements.

ABSTRACT. Recent deep learning-based video object tracking approaches have shown promising performance on standard tracking benchmarks. However, the scarcity of low-light scenes in common tracking datasets poses a significant challenge for developing robust low-light deep trackers, especially for the Unmanned Aerial Vehicle (UAV) tracking scenario with more difficulty for data collection, thus severely degrading the performance of existing deep trackers in low-light UAV conditions. In this paper, we propose a simple yet effective approach to enhance UAV tracking in low-light conditions. Different from previous works that need annotated pairwise images for lightness enhancement, our proposed approach is annotation free, which can be generally used to improve existing trackers in low-light scenario. Specifically, we first simulate low-light conditions by performing simple data augmentation on normal tracking videos. We then conduct the feature alignment between the normal and low-light video frames, in order to facilitate the trackers to adapt well in low-light conditions. Surprisingly, we find that our simple approach can effectively improve existing deep trackers, including both typical CNN-based SiamFC and transformer-based OSTrack. Experiments on UAVDark135 show that our proposed trackers can achieve favorable performance with simple design, which has the potential to be served as a simple yet effective baseline in low-light UAV tracking.

ABSTRACT. It is a hot issue to discuss the robust control policy for automated manufacturing systems (AMSs) with unreliable resources in recent years. Before designing a robust controller, it is necessary to analyze the production processes of AMSs with unreliable resources. In our work, we study the classification of production states for AMSs with multiple unreliable resources. First, Petri nets are utilized to model the above systems. Second, the reachable markings of Petri nets describing undesired production states are classified into four types, such as deadlocks, F-deadlocks, F-blockings and FP-blockings by the structural characterization, maximal perfect resource-transition circuits, of Petri nets. Next, we prove that all undesired reachable markings are only the above four types for a kind of AMSs with multiple unreliable resources. Finally, an algorithm to compute all deadlocks and blockings with low computational- complexity is presented, and an example is utilized to illustrate the validity of the conclusion and the algorithm.

ABSTRACT. With the continuous development of social economy, artificial intelligence technology has gradually become an indispensable part of various industries, and convolutional neural network (CNN) shows its powerful learning ability in the field of medical image diagnosis. In the traditional medical field, the amount of medical image data is huge, and doctors need to go through long-term professional training and rich practical experience in order to interpret it, which is a time-consuming and labour-intensive process. In contrast, convolutional neural network, with their powerful learning computational ability, can effectively handle large amounts of medical image data. This work aims to investigate the convolutional neural network algorithm to identify image data of malignant and benign skin cancers. In addition, in order to further improve its classification accuracy, a novel attention mechanism, namely input attention mechanism, is integrated into it, and the convolutional neural network model with the proposed attention mechanism (CNN-IAM) is proposed. Finally, the effectiveness of the proposed model is verified on a large amount of skin image data.

ABSTRACT. In recent years, cyberbullying has become increasingly severe, causing significant impacts on both individuals and society. Textual information on social media plays a vital role in studying cyberbullying behaviors. Therefore, extracting and understanding the important part heavily affects the detecting performance. While evaluating mainstream word embedding models for cyberbullying detection, we propose a Neural Interaction Networks by Comment Embedding (NINCE) model to address missing positional and segment information in comment encoding. After comparing it with four different algorithms across three datasets, the experimental results demonstrate the superiority of the proposed model in terms of both accuracy and efficiency.

ABSTRACT. Rod pumping systems are instrumental in the extraction of oil. Nevertheless, the complex operational environment increases their likelihood of malfunctions, thereby impacting both oil yield and the longevity of the equipment. To address the potential faults that may emerge within a rod pumping system, we propose a multi-model ensemble approach for fault diagnosis. The method utilizes models with 1 to 8 Encoder modules as base learners to extract deep-level features from the displacement and load condition data of the rod pumping system from different perspectives. Finally, the Dempster-Shafer(D-S) evidence theory is used in the decision-making component to integrate various feature information gleaned by the various models, facilitating a comprehensive fault diagnosis for the rod pumping system. In order to substantiate the efficacy of our proposed method, it is compared with traditional machine learning methods such as logistic regression and random forests, as well as Convolutional Neural Network(CNN) methods for images. A total of 8878 actual production operation data points are used for training and testing. Our method demonstrates a remarkable accuracy rate of 97.82%. This not only underscores its superior generalization capabilities but also highlights its capacity to significantly mitigate computational demands, offering good practical value.

ABSTRACT. Magnetically responsive porous materials haveemerged as promising candidates for advanced, stimulus-drivendrug delivery systems. This study introduces a cost-effective3D printing technique for fabricating flexible porous scaffoldsembedded with NdFeB microparticles, giving magnetic sensitivityto the materials. These multifunctional scaffolds feature ad-justable porosity, making them versatile for both large-scale andpersonalized drug delivery applications. Through experimentalevaluation, we demonstrate that the designed magnetically softporous scaffolds exhibit exceptional flexibility, returning to theiroriginal shape after deformation and achieving a compressionrate of 49 ± 0.7% under the influence of N52-grade magnets

ABSTRACT. With the rapid development of artificial intelligence and neural networks, deep reinforcement learning has achieved remarkable results in a series of complex sequential decision-making problems. The application of multi-agent reinforcement learning in air combat game scenarios is also booming. In the use of reinforcement learning for multi-agent air combat decision-making, the scalability and transferability of the model have become critical issues. Designing a multi-agent air combat decision-making framework with solid scalability, robustness, and rapid convergence has become a research hotspot in various countries. To address this problem, this paper proposes a multi-agent air combat decision-making framework based on attention mechanism transfer and designs a 2D air combat simulation environment for this framework. The decision-making process of this framework is divided into two stages. First, course learning is carried out in the designed essential air combat environment to enhance the aircraft's combat capability. Then, the trained strategy is transferred to a complex air combat environment for further training. Experiments have shown that this framework has better transferability and robustness.

ABSTRACT. With the advancement of warehousing technology, the allocation of multiple tasks in multi-AGV systems has become an important factor in order to decrease task ompletion time in the warehouse. This paper develops a multi-task scheduling algorithm that employs a dual-task chain approach to maximize task execution benefits. For the multi-task allocation problem, this paper proposes a dual-task chain-based task allocation algorithm (DTTA), which uses the dual-task chain strategy as a mechanism for selecting multiple tasks, taking into consideration derived tasks to enhance solution quality. Simulation experiments show that the proposed algorithm and model perform better thanother methods.

ABSTRACT. Vehicle congestion or collision can easily happen at unsignalized intersections if there are no appropriate control policies. In this paper, we mainly focus on special congestions due to vehicle circular waits, i.e., deadlocks. Based on Petri nets, a small-size deadlock prevention policy is designed to ensure that vehicles can pass smoothly and efficiently through any unsignalized intersection. Firstly, the two-way four-lane unsignalized intersection is discretized and the system is modeled as S3PRs. Secondly, an effective transition cover of the system is computed and then a small-sized Petri net controller based on the effective transition cover is synthesized. Finally, the liveness of the proposed controller is proved, which shows that the smooth traffic at the unsignalized intersection can be guaranteed. That is, the effectiveness of the deadlock prevention policy is shown. Meanwhile, its advantage is also illustrated by comparing with methods in literatures.

ABSTRACT. Artificial intelligence has advanced quickly, making it possible to analyze brain pictures for tumor detection. Traditional computer vision techniques, while capable of identifying tumors, often lack the precision and adaptability required for accurate pre-diagnosis. When combined with deep learning, this has improved the accuracy of tumor identification and pre-diagnosis by determining the position and size of the tumors. This advancement is helpful in guaranteeing that sufferers receive prompt treatment. Consequently, enhancing learning accuracy has emerged as a foundational necessity in the realm of medical image diagnosis. Inspired by dendritic neurons, researchers have devised the dendritic neuron model (DNM), which emulates the information processing characteristics of brain neurons within neural circuits. Combining this neuron with traditional deep learning models has become widely popular and consistently yields excellent results in solving classification problems. In this paper, we propose a novel architecture called DMobileNet, which by leveraging the strengths of MobileNet’s lightweight design and DNM’s capability to emulate dendritic neuron behavior, DMobileNet achieves dynamic synaptic connections and enhanced information processing capabilities. Experimental results across various tasks demonstrate that DMobileNet outperforms both traditional MobileNet and established deep learning models in terms of accuracy and computational efficiency. In the case of the brain tumor problem, our model achieves an accuracy of 97.4\% and an F1 score of 96.9\% in F1 score. For classification tasks, this study proposes that utilizing DNM as a classifier may facilitate the advancement of more effective deep learning models.

ABSTRACT. With the development of Internet of Things (IoT), various Internet of Things devices (IoTDs) are widely applied to bestow modern society efficiency and convenience. Nowadays, IoTDs are generally equipped with computational capabilities to process the sensed data. However, performing computations on IoTDs consumes a significant amount of energy at the IoTDs. In recent years, mobile edge computing (MEC) can partake the IoTDs' computational tasks and effectively reduce the energy consumption of the IoTDs' local computation. Meanwhile, it is still inconvenient to power such a tremendous number IoTDs via traditional energy supply methods. Recently, far-filed wireless charging is a promising technique to remotely power surrounding IoTDs. Yet, there are few studies that jointly consider the MEC technique and the green energy-powered far-field wireless charging technique to improve the survival percentage of IoTD. Therefore, this paper proposes a MEC and green energy far-field wireless charging facilitated network to reduce the IoTDs' energy consumption and further provision energy supply to them. We first propose the energy consumption models of the IoTD, the Mobile Charging Vehicle (MCV) and the base station, and the energy replenishment model of the IoTD. We then formulate the maximum IoTDs' survival percentage problem to enable more IoTDs' of being alive. As this IoTDs survival percentage maximization problem is proved to be NP-hard, and building on our use of MEC technology to reduce energy consumption, we further propose the Least Energy Loss (LEL) algorithm to decrease the number of MCV's anchors points (APs) to convey more energy to the IoTDs. Extensive simulations validate that our proposed algorithm can effectively improve the IoTDs' survival percentage.