Keywords: Deep Learning, Neural Networks and their Applications, Application of Artificial Intelligence
Abstract: CTR model has played an important role in modern recommendation systems. Most of the recommendation models in industrial scenario are trained by TensorFlow. However, we observed that, TFRecord, the native k-v data format in TensorFlow, is not the best choice for CTR training. Those keys take up to 54% of the storage space in TFRecord formatted training data. To overcome this, we introduce Apache Parquet, a column-oriented data format, into CTR tasks to improve spatial efficiency. Besides, to use in production environment, we further give some high performance implementations of Parquet training scheme. Firstly, GPU data preprocessing method is adopted in replace of original Spark based solution to generate Parquet training data and accelerate data preprocessing. Secondly, we modify data loader in TensorFlow to consume the Parquet training data with high efficiency. Experimental results show that, the size of preprocessed Criteo dataset is 95.29% smaller in comparison to TFRecord and the data preprocessing time also reduces 99.6%. Without any model performance damage, we speed up the training process by 1.45x. Our scheme has applied to our internal business and has obtained similar performance benefits.

Keywords: Neural Networks and their Applications, AI and Applications, Application of Artificial Intelligence
Abstract: To enhance the accuracy of tidal bore height prediction in the Qiantang River, this paper addresses the limitations of one-dimensional, single-scale time series in terms of representational capacity by proposing an MPCSM model that integrates multi-period decomposition with cross-scale fusion for tide height forecasting. The model initially employs the Maximum Overlap Discrete Wavelet Transform to decompose the original time series data into multiple periodic components. Subsequently, it leverages multi-frequency channel attention and gating mechanisms to ascertain the significance of each frequency component and optimize prediction weights. For prediction, cross-scale iterative forecasting is utilized to capture both long-term and short-term characteristics of tidal height data, accompanied by a designed loss function computation strategy that adapts to cross-scale prediction errors. At the Qiantang River Zhakou station, compared to current mainstream prediction models, the average absolute error across different forecast time spans has been reduced by 9.26%. The research findings can serve as a reference for tidal bore prediction in the Qiantang River and other similar basins.

Keywords: Neural Networks and their Applications, AI and Applications, Deep Learning
Abstract: Session-based recommendation (SBR) is a challenging task that aims to make item recommendations based on anonymized user session data. Mainstream SBR efforts focus on modeling information within a session and do not use information from other sessions. Although some works try to use other session information, there are still many limitations, and how to model other session information is still a challenging task. To overcome these limitations, we propose a new method for learning similar intentions between sessions, aiming to better model the recommendation information contained in other sessions. Specifically, we contribute a new model named ISIM-GNN that learns and integrates three levels of information simultaneously: (i) In the intra-session representation learning layer, we represent the session as a session graph and model it using a gated graph neural network. (ii) In the global item embedding learning layer, We use the graph attention mechanism to propagate and aggregate relevant item information from other sessions on the global graph. (iii) In the inter-session similar intent learning layer, we employ both "hard similarity" and "soft similarity" to select similar sessions, and use the attention mechanism to conduct session-level aggregation on the selected similar sessions to make better use of the inter-session collaboration information. Experiments on three real-world datasets show a significant performance improvement of our approach compared to state-of-the-art work.

Keywords: Deep Learning, Transfer Learning, Machine Learning
Abstract: It has been acknowledged that the risks posed by Advanced Persistent Threats (APTs) are critical. These attacks can allow cybercriminals to remotely manipulate infected devices and steal sensitive data. To effectively combat APT attacks, it is crucial to employ multidimensional analysis techniques that can predict their impact and detect lateral infiltration behavior. This paper presents an approach called AutoASD, which utilizes Neural Architecture Search (NAS) Deep learning (DL) and Transfer Learning (TL) to identify various types of malicious traffic in APT attacks. AutoASD analyzes data at multiple granularities to classify different types of malware traffic and enhance classification accuracy. It leverages feature extraction and a pre-trained high-performance backbone network as the seed network, and employs parameter remapping to adjust the depth, width and kernel to create a super network. The aim of using NAS is to improve the real-time and accuracy of the system. In experiments, the effectiveness of AutoASD was verified using MobileNetV2 and demonstrated superior performance in APT malicious traffic classification, particularly for attacks with small sample sizes.

Keywords: Knowledge Acquisition, Neural Networks and their Applications, Deep Learning
Abstract: Continual Learning for Named Entity Recognition (CL-NER) is designed to train models capable of adapting to evolving data by continuously introducing new entity types. This approach is crucial in dynamic environments where data evolves, such as social media, healthcare, and legal documents, necessitating the model to retain the memory of previously learned entity types while learning to identify new ones. However, due to the neural network's tendency to acquire new knowledge and forget old knowledge in continual learning and the unique non-entity type annotations in NER tasks, CL-NER faces severe catastrophic forgetting and semantic drift issues. In this paper, we propose Enhanced Knowledge Distillation and Entropy-based Adaptive Pseudo Label Weights (EDAW) to address the catastrophic forgetting and semantic drift issues in CL-NER. Specifically, we develop an enhanced knowledge distillation method that combines Kullback-Leibler divergence and feature cosine discrepancy. This method effectively minimizes the variance in output probability distributions and aligns the internal feature spaces between new and old models, thus reducing catastrophic forgetting. Additionally, we propose an entropy-based adaptive pseudo label weight method that allows the model to assign different weights to pseudo labels with varying certainties during training, effectively alleviating semantic drift and error accumulation caused by erroneous re-labeling of pseudo labels. Notably, this study pioneers the inclusion of a Chinese dataset in CL-NER, enhancing the model's robustness and demonstrating its efficacy in a multilingual context. Experiments on fourteen CL-NER settings across four public NER datasets show that EDAW improves average Micro-F1 and Macro-F1 scores by 3.44% and 3.72%, respectively, over existing state-of-the-art(SOTA) methods. We make our code available at: https://github.com/livosr/EDAW/tree/master

Keywords: Multimedia Computation, Neural Networks and their Applications, Computational Intelligence
Abstract: Neural architecture search (NAS) automates neural network design by using optimization algorithms to navigate architecture spaces, reducing the burden of manual architecture design. While NAS has achieved success, applying it to emerging domains, such as analyzing unstructured 3D point clouds, remains underexplored due to the data lying in non-Euclidean spaces, unlike images. This paper presents Success-History-based Self-adaptive Differential Evolution with a Joint Point Interaction Dimension Search (SHSADE-PIDS), an evolutionary NAS framework that encodes discrete deep neural network architectures to continuous spaces and performs searches in the continuous spaces for efficient point cloud neural architectures. Comprehensive experiments on challenging 3D segmentation and classification benchmarks demonstrate SHSADE-PIDS's capabilities. It discovered highly efficient architectures with higher accuracy, significantly advancing prior NAS techniques. For segmentation on SemanticKITTI, SHSADE-PIDS attained 64.51% mean IoU using only 0.55M parameters and 4.5GMACs, reducing overhead by over 22-26X versus other top methods. For ModelNet40 classification, it achieved 93.4% accuracy with just 1.31M parameters, surpassing larger models. SHSADE-PIDS provided valuable insights into bridging evolutionary algorithms with neural architecture optimization, particularly for emerging frontiers like point cloud learning.

Keywords: Human-Collaborative Robotics, Design Methods, Assistive Technology
Abstract: As labor shortage is rising at an alarming rate, it is imperative to enable all people to work, particularly people with disabilities and elderly people. Robots are often used as universal tool to assist people with disabilities. However, for such human-robot workstations universal design fails. We mitigate the challenges of selecting an individualized set of input and output devices by matching devices required by the work process and individual disabilities adhering to the Convention on the Rights of Persons with Disabilities passed by the United Nations. The objective is to facilitate economically viable workstations with just the required devices, hence, lowering overall cost of corporate inclusion and during redesign of workplaces. Our work focuses on developing an efficient approach to filter input and output devices based on a person's disabilities, resulting in a tailored list of usable devices. The methodology enables an automated assessment of devices compatible with specific disabilities defined in International Classification of Functioning, Disability and Health. In a mock-up, we showcase the synthesis of input and output devices from disabilities, thereby providing a practical tool for selecting devices for individuals with disabilities.

Keywords: Human-Collaborative Robotics, Design Methods
Abstract: While automation is widespread in tailoring, high- end fashion brands still rely on skilled manual labor for intri- cate stitching. However, finding skilled workers is challenging due to a lack of new talent entering the field. The stitches required both speed and precision. In response to the declining availability of skilled artisans, this study explores leveraging cobot technology to replicate traditional manual techniques. The research investigates the development and implementation of a system integrating cobots to execute precise stitching tasks. Key aspects include designing infrastructure to support cobots and implementing advanced techniques for fabric manipulation and needle guidance. The study also examines the adaptation of cobot technology to replicate the ”Ladder Stitch” technique, traditionally performed by skilled artisans. By blending tra- ditional craftsmanship with modern technology, this research aims to address the shortage of skilled labor in the tailoring industry.

Keywords: Human-Collaborative Robotics, Assistive Technology, Shared Control
Abstract: The implicit assumption that human and autonomous agents have certain capabilities is omnipresent in modern teaming concepts. However, none formalize these capabilities in a flexible and quantifiable way. In this paper, we propose Capability Deltas, which establish a quantifiable source to craft autonomous assistance systems in which one agent takes the leader and the other the supporter role. We deduct the quantification of human capabilities based on an established assessment and documentation procedure from occupational inclusion of people with disabilities. This allows us to quantify the delta, or gap, between a team’s current capability and a requirement established by a work process. The concept is then extended to the multi-dimensional capability space, which then allows to formalize compensation behavior and assess required actions by the autonomous agent.

Keywords: Intelligent Transportation Systems, Autonomous Vehicle, Robotic Systems
Abstract: This paper presents a novel trajectory planning approach for two typical aerial transportation systems: UAV-slung-load and flying inverted pendulum. By integrating Rapidly-exploring Random Trees* (RRT*) into Nonlinear Model Predictive Control (NMPC), the proposed method enhances motion planning, enabling effective navigation in complex environments while ensuring stability and safety. Simulation results demonstrate the approach's capability to overcome local minima and generate feasible trajectories, highlighting its potential to advance trajectory planning in UAV transportation systems.

Keywords: Intelligent Transportation Systems, Consumer and Industrial Applications, Manufacturing Automation and Systems
Abstract: In the era of rapidly advancing technology, food classification has emerged as a pivotal application across various domains including health monitoring, dietary assessment, and culinary innovation. However, efficiently categorizing food items remains a challenge, particularly in scenarios with limited labeled data. This paper introduces a novel approach for few-shot food classification using Prototypical Networks with ResNet-50 and an attention mechanism as embedding network. Leveraging the inherent capability of Prototypical Networks to learn from scarce examples, our method demonstrates exceptional adaptability and accuracy in classifying food items. Through extensive experimentation on the Food-101 dataset, employing various CNN architectures, our findings underscore the effectiveness of our approach. In particular, ResNet-50 integrated with the attention mechanism surpasses other architectures, achieving superior classification accuracies of 91.5% and 95.2% for 1-shot and 5-shot learning scenarios, respectively. This integrated approach showcases the potential of Prototypical Networks in addressing the challenges of limited labeled data in food classification tasks, marking a significant advancement in the field.

Keywords: Intelligent Transportation Systems, Decision Support Systems, System Modeling and Control
Abstract: Feature selection holds significant importance in knowledge mining as it plays a pivotal role in selecting and preserving the most informative features within a dataset while discarding irrelevant, redundant, or noisy attributes. This process contributes to enhancing model performance, reducing computational complexity, and refining interpretability, thus facilitating more accurate and efficient data analysis. In high-dimensional datasets, the necessity for feature selection becomes more pronounced due to the heightened risk of encountering the curse of dimensionality. Therefore, this study proposes a weighted fuzzy rough quickreduct (FRQR) feature selection approach employing feature weights to handle the equal situation problem inherent in FRQR. The proposed method is evaluated on ten publicly available datasets with feature sizes ranging from 2000 to 15154. The selected features are used to train and test random forest candidate models whose estimates are then combined according to the posterior probabilities by Bayesian Model Averaging (BMA). The performance of the model on the selected features is evaluated on four performance metrics. The essentiality of the selected features is further determined by comparing the model classification performance achieved on the non-selected features and all the dataset features. The results indicate competitiveness in the performance metric values achieved on selected features over the other two feature categories affirming the efficacy of the proposed method.

Keywords: Autonomous Vehicle, Decision Support Systems, Intelligent Transportation Systems
Abstract: In order to solve the artificial problems such as high prices and frequent conflicts between drivers and consumers in ride-hailing services, China and other countries are actively promoting the development of shared autonomous driving. We explore the stimulative impact of subsidy initiatives on shared autonomous driving, focusing on consumers' green preferences. By constructing a Hotelling model of AV and HV service platform competition, it analyzes the optimal solutions under three scenarios: no subsidy, subsidized AV platform, and subsidized consumers, and conducts sensitivity analysis. The study found that subsidizing AV platforms can reduce their service prices and enhance market competitiveness, while subsidizing consumers may increase their service prices but not necessarily increase market share. Policy makers should adjust subsidies as AV services develop, and may gradually cancel incentives after the industry matures. At this time, AV platforms need to improve their technology and management level to achieve sustainable development.

Keywords: Autonomous Vehicle, Intelligent Transportation Systems, Fault Monitoring and Diagnosis
Abstract: Online corner case detection is crucial for ensuring safety in autonomous driving vehicles. Current autonomous driving approaches can be categorized into modular approaches and end-to-end approaches. To leverage the advantages of both, we propose a method for online corner case detection that integrates an end-to-end approach into a modular system. The modular system takes over the primary driving task and the end-to-end network runs in parallel as a secondary one, the disagreement between the systems is then used for corner case detection. We implement this method on a real vehicle and evaluate it qualitatively. Our results demonstrate that end-to-end networks, known for their superior situational awareness, as secondary driving systems, can effectively contribute to corner case detection. These findings suggest that such an approach holds potential for enhancing the safety of autonomous vehicles.

Keywords: Autonomous Vehicle, Intelligent Transportation Systems
Abstract: With the continuous advances of technologies, the demand for precise vehicle positioning has grown significantly. Although it is now possible to capture driving scenes and record driving paths using a dashcam, standard civilian GPS typically has the accuracy errors ranging from 3 to 5 meters. This level of precision is in general not sufficient for the rapidly evolving ADAS (Advanced Driver Assistance Systems). In this paper, we present a high-precision vehicle positioning technique based on satellite map and image data. Instead of using expensive LiDAR sensors, the proposed approach utilizes lane detection, semantic segmentation and geolocation to extract environmental features from images. A* algorithm is then adopted to refine the driving trajectory for the improvement of vehicle positioning accuracy. Furthermore, we establish an image dataset containing satellite maps and latitude/longitude coordinate information of various road scenes. The code and datasets are made available publicly at https://github.com/M610415018/M610415018-Paper.

Keywords: AIoT, Application of Artificial Intelligence, Machine Learning
Abstract: 随着大数据、微处理器和其他应用的进步，物联网 （IoT） 取得了长足的发展。由于缺乏必要的安全防御机制，物联网设备容易被攻击者针对和控制。他们可以操纵大量的物联网设备，对一个国家或地区的网络基础设施发动DDoS攻击，导致严重的经济损失和社会安全风险。基于深度学习的入侵检测方法通常依赖于大量高质量的训练实例，因此很难将其应用于缺乏足够标记数据的网络流量。传统的机器学习 （ML） 方法在从高维数据中提取和表示特征方面的能力有限，这使得发现数据中的底层结构和模式变得具有挑战性。针对上述问题，该文提出一种融合了快速独立分量分析（FastICA）模块和多粒度级联森林（GcForest）的入侵检测系统（IDS）。通过利用 Fa

Keywords: Representation Learning, Deep Learning, Machine Learning
Abstract: Model pruning, as an effective method for compressing large language models (LLMs), has recently attracted considerable attention in the field of natural language processing. However, existing LLM pruning methods have two main drawbacks: (1) Iterative pruning for LLMs with over a billion parameters requires retraining, which leads to significant pruning costs. (2) LLMs Pruning is formalized as a weight reconstruction problem that necessitates second-order information, incurring expensive computations. To address these issues, we propose a novel pruning method named Eplra: efficient one-shot pruning of large language models with low-rank approximation, which efficiently identifies sparse networks in LLMs. Specifically, we design a novel pruning metric based on input activations for the rapid one-shot compression of LLMs. We first incorporate input activations into the calculation of weight importance to promote precise pruning of low-priority weights. Then, we perform local weight comparisons across each output of linear layers to induce uniform sparsity. Next, we expand Eplra into semi-structured pruning patterns to accommodate various acceleration scenarios. Finally, we employ low-rank parametrized update matrices to fine-tune the pruned model, facilitating a swift recovery of model performance. Experimental results on various language benchmark datasets demonstrate that Eplra outperforms the state-of-the-art methods.

Keywords: Image Processing and Pattern Recognition, Neural Networks and their Applications, Deep Learning
Abstract: With rapid technological advancements in computer vision, the recognition of abnormal behavior during examinations has transitioned from human observation to computer-assisted recognition. Although traditional 2D Convolutional Neural Networks (CNNs) excel in computational efficiency, they need to capture crucial temporal dynamics for comprehensive video analysis more precisely. Nevertheless, 3D CNN-based methods demonstrate promising performance in temporal modeling but impose substantial computational demands and deployment costs. To overcome these challenges, this paper introduces an innovative Examination Anomaly Action Recognition Network named ReTANet. It incorporates cross-channel temporal modeling to capture temporal features within videos. It also employs Multi-Scale Channel Attention to enrich feature representation and extract channel and spatial information, thereby enhancing recognition accuracy without significantly increasing computational complexity and model parameters. Furthermore, this paper introduces the Examination Anomaly Action Dataset, also named the ExamGuard Dataset (EGD), to facilitate model training and evaluation. Remarkably, our model demonstrates superior performance compared to existing mainstream action recognition algorithms on the HMDB-51 dataset. Rigorous ablation studies conducted on the UCF-101 dataset have shown the effectiveness and significance of the proposed module.

Keywords: AI and Applications, Application of Artificial Intelligence, Deep Learning
Abstract: Extracting entity relations is vital in legal artificial intelligence. It automates the mining of triple data from vast legal texts. Current methods face challenges in inaccurately identifying legal named entity boundaries and extracting overlapping relation triples from legal texts. We present KGNet, a model developed to address these issues effectively. Our approach introduces a Word Information Generator Based on BMES tagging combined with the Fusionformer module. This innovation enhances the incorporation of legal domain knowledge into text representations, improving the accuracy of entity recognition. Additionally, we utilize the GlobalPointer decoder, which redefines and decomposes relation triples, thus resolving the issue of overlapping entities. Performance evaluations on a specially constructed judicial document dataset show that KGNet achieves an F1 score of 66.7%, representing an average improvement of 15.3% over baseline models. These results confirm the effectiveness of KGNet in enhancing legal document processing.

Keywords: Adaptive Systems, Cooperative Systems and Control
Abstract: 消防无人机技术已成为其中之一 现代消防行动的核心工具，及其 任务 执行的规模和复杂性都在不断扩大。 面对这样的发展，更是变得更加 找到一种有效的方法来确保无人机能够 被指派执行最合适的任务。在这个 研究中，我们使用了 Environment-Class、Agent、Role、Group、 和对象（E-CARGO）模型，以系统地分析 消防无人机的任务分配（FDTA）问题，以及 引入了增强的鲸鱼优化算法 （EWOA） 优化FDTA问题中的路径规划。最后 仿真实验在多样化下进行 地形 展示效率和快速响应的条件 改进算法在不同工作负载下的能力 和环境条件。

Keywords: Image Processing and Pattern Recognition, Machine Learning, Deep Learning
Abstract: Vision Transformers (ViTs) have achieved excellent performance on many computer vision tasks, which has attracted attention of many researchers for their adversarial robustness. As a kind of black-box attack, transfer-based attacks usually use adversarial examples generated by a surrogate model to attack structurally different models. It is practical and poses a certain threat to the application of ViTs in critical security areas. Existing transfer-based attacks against ViTs suffer from weak adversarial transferability and noticeable perceptibility. In this work, we propose a method called Reduce Regional Perturbation Interaction and Differentiated (RPID) attack, which employs two strategies of reducing correlation between regional perturbations and adding differentiated perturbations to produce adversarial examples. Extensive experiments demonstrate that our proposed method improves the transferability of the baseline methods for adversarial attacks against ViTs while maintaining stealthiness.

Keywords: Knowledge Acquisition, Representation Learning, Neural Networks and their Applications
Abstract: Knowledge graphs (KGs) provide a structured representation of the real world through entity-relation triples. However, current KGs are often incomplete, typically containing only a small fraction of all possible facts. This involves inferring missing content from existing information, a task known as link prediction. Existing methods in the field of link prediction struggle with controlling the dimensionality of embedding vectors or suffer from overly complex models. In order to tackle these challenges, we introduce a method in this paper, named Low-Dimensional Embedding Method for Link Prediction Combining Self-Attention and Enhanced-TuckER (LESaET). LESaET leverages both self-attention mechanisms and tensor factorization to learn expressive contextual-enhanced representations of KGs. Specifically, LESaET employs the multi-head self-attention mechanism of Transformer as an encoder to capture the mutual information between entities and relationships, and utilizes Enhanced-TuckER as a decoder, ultimately achieving expressive low-dimensional embeddings for link prediction tasks. LESaET demonstrates competitive performance compared to advanced methods on standard datasets.

Keywords: Multimedia Computation, Deep Learning, Media Computing
Abstract: In this study, we address the issue of adversarial robustness within the context of Blind Image Quality Assessment (BIQA), an area of heightened importance due to the inherent susceptibility of Deep Neural Networks (DNNs) to adversarial assaults. Current approaches primarily rely on adversarial training, which, despite its efficacy, imposes a significant computational burden. Our research proposes an alternative strategy known as the Soft Thresholding Norm (ST Norm). This approach counters the 'feature shift' phenomenon, identified by a substantial Euclidean Distance Statistics (EDS) between original and adversarial features, through the imposition of sparse constraints on potential features following batch normalization. This novel method offers several advantages: it reduces the Lipschitz constant yielding smoother models, seamlessly integrates with existing models, and boasts inherent denoising capabilities, thereby effectively mitigating the impact of adversarial perturbations. Results suggest that our approach achieves robustness comparable to adversarial training but with significantly less computational overhead. Moreover, it consistently outperforms other adversarial defense strategies on BIQA datasets, highlighting its practical effectiveness in enhancing adversarial robustness. This study underscores the potential of the Soft Thresholding Norm within the realm of IQA tasks, positioning it as a resource-efficient alternative to traditional adversarial training methodologies.

Keywords: Application of Artificial Intelligence, Knowledge Acquisition
Abstract: The NER task in manufacturing is usually lack sufficient labeled data resources. To tackle this issue, this paper presents an effective NN-PLM framework for few-shot NER in manufacturing, which introduce a simple enhancement of the prompt-based learning model using nearest neighbor retrieval. We retrieve the morphologically similar characters for each character to be predicted and then rectifies the prediction. Moreover, we use supervised contrastive learning (SCL) and instance weighting to get better semantic representations of multi-category characters. Compared with the best baseline, our NN-PLM achieves a 7.12% F1 score average improvement on all few-shot settings in manufacturing.

Keywords: Deep Learning, Neural Networks and their Applications, Expert and Knowledge-Based Systems
Abstract: Language Models (LMs) have achieved impressive success in various question answering (QA) tasks but have shown limited performance on structured reasoning. Recent research suggests that Knowledge Graph (KG) can augment text data by providing a structured background to enhance reasoning capabilities of LMs. Therefore, how to integrate and reason over KG representations and language context remains an open question. In this work, we propose MJR, a novel model to integrate encoded representations of LMs and graph neural network through multiple layers of feature interaction operations. Subsequently, the fused feature representations in two modalities are fed into a multi-head representation fusion module to comprehensively capture semantic and graph structure information, thereby enhancing language understanding and reasoning capabilities. In addition, we investigate the performance and applicability of different types of large language models as text encoder in the question-answering task. We evaluate our model on three common dataset: CommonsenseQA, OpenBookQA, and MedQA-USMLE datasets. The results demonstrate the advancements of MJR over existing LMs, LM+KG and LLMs models in reasoning for question answering.

Keywords: Application of Artificial Intelligence, Deep Learning, Machine Learning
Abstract: Friendly robots have extremely important application prospects in many fields. However, in unstructured environment, the interaction between manipulator and dynamic environments faces the problems of high uncertainty caused by random invasion of work space and computational complexity brought by multi-dimensional action space. Therefore, we propose a hierarchical planning algorithm based on global path guidance reinforcement learning to solve this problem from the decision and planning level. Specifically, the global path planning algorithm first produces a global reference path that ensures the target can be reached. Then the reference path is decomposed into consecutive local targets, which are combined with the objective function of reinforcement learning as local constraints. Finally, the reinforcement learning local planner generates the action of the manipulator based on the observed information. The simulation results show that our method is superior to the standard off-policy reinforcement learning algorithm in terms of learning speed and accuracy, which proves the effectiveness of our algorithm.

Keywords: Application of Artificial Intelligence, Deep Learning, Machine Learning
Abstract: Deep Neural Networks (DNNs) are highly susceptible to adversarial noise, which can lead to erroneous predictions. In high-stakes scenarios, such as autonomous driving and medical diagnosis, DNNs inaccuracies can be dire. To address this issue, Adversarial Training (AT) has been widely adopted as an effective defense method. However, our analysis reveals two critical flaws in the traditional AT approach that hinder its adversarial robustness: (1) focus only on a subset of robust features during the training process. This narrow focus limits the model's ability to learn and perceive a diverse range of features. (2) tend to overlook potential cues in non-robust features that could be beneficial for the model to make correct predictions. These cues, referred to as "positive activations" for simplicity, contain valuable information that can enhance the model's perception and understanding of the input data. In this way, we propose a novel and plug-and-play framework called Multi-branch Feature Fusion and Dynamic Reconstruction (MFFDR), which leverages multi-branch attention mechanisms to enhance the model's perception of robust features and enrich the diversity of learned features. Moreover, we employ a dynamic weighting strategy to reconstruct non-robust features in order to utilize the positive activations embedded within them. Extensive experiments demonstrate that our method significantly improves the model's adversarial robustness and outperforms previous state-of-the-art methods.

Keywords: Application of Artificial Intelligence, AI and Applications, Neural Networks and their Applications
Abstract: With the extensive application of network traffic encryption technology, the accurate and efficient classification of encrypted traffic has become a critical need for network management. Deep learning has become the predominant method for traffic classification, primarily involving the transformation of network traffic into grayscale images and their subsequent classification using Convolutional Neural Networks (CNNs). However, traditional grayscale image generation methods are plagued with issues of redundant and lost information, and conventional channel attention mechanisms are still insufficient in capturing key traffic features, collectively hindering the enhancement of classification performance. To tackle these issues, this paper introduces a classification method based on Byte Transfer Probability and Coordinate Attention Mechanism in Residual Network (BTP-CAResNet). This method, on the foundation of the classic ResNet architecture, incorporates a new grayscale image generation method that utilizes Byte Transfer Probability, effectively overcoming the deficiencies of traditional approaches. Additionally, this paper integrates a Coordinate Attention Mechanism into the ResNet model, which effectively overcomes the limitations of traditional channel attention mechanisms and further improves the performance of traffic classification. Experimental validation on the ISCX VPN-nonVPN dataset demonstrates that, compared to previous CNN-based methods, the method proposed in this paper exhibits superior performance in key metrics such as accuracy, precision, recall, and F1 score. It provides a new perspective for traffic classification based on convolutional neural networks.

Keywords: AI and Applications, Big Data Computing,, Deep Learning
Abstract: Knowledge Graph Completion (KGC) aims to enrich and complete the knowledge graph by discovering missing information from existing fact triples. However, existing KGC methods often overlook the utilization of structured knowledge within the knowledge base. In this paper, we propose a novel Large Language Models-based Knowledge Graph Completion framework, called SIKGC, which builds the structural information prompt to assist the knowledge graph completion tasks. Specifically, we arrange the triples in the knowledge graph as the sequences of text. By fusing the descriptions of entities, relations and their structural information as task-aware prompts, we input such prompts into large language models and regard the responses as prediction tasks. The experimental results on various public datasets show that the proposed method outperforms all baseline methods for the three knowledge completion tasks and attains state-of-the-art in triple classification. We also demonstrate that fine-tuning the smaller large language models (e.g., Baichuan2-13B, LLaMA2-13B, ChatGLM3-6B) with relevant data markedly enhances their KGC capabilities and significantly outperforms GPT-4.

Keywords: Big Data Computing,, Machine Learning
Abstract: Concept drift refers to the probability distribution of data generation changes over time in a data stream environment. In recent years, there has been an increasing interest in drift detection models. However, due to the lack of labeled concept drift datasets, most researchers tend to using synthetic drift data generators for model training. These generators only have relatively simple feature distributions, which fail to capture the complexity found in real-world scenarios. This paper introduces a real scenario concept drift label generator (RealDriftGenerator). This generator aims to preserve the complexity and temporal correlation of real-world scenario while generating concept drifts with user defined drift positions and drift widths. The validation results shows that the temporal correlation coefficients of RealDriftGenerator is significantly higher than benchmark drift generators. Additionally, the ability of RealDriftGenerator to capture the complexity in real-world scenarios is 20% higher than benchmark drift generators(measured by model performance). The source code of RealDriftGenerator has been published on https://github.com/sniperrifle71/realDriftGenerator.

Keywords: Agent-Based Modeling, Artificial Social Intelligence
Abstract: Opinion dynamics studies the principles governing the evolution of collective opinion, offering valuable insights into the comprehension of social phenomena and forecasting group behavior. However, existing opinion dynamics models often overlook the impact of both opinion climate and cognitive capacities on interactive behaviors, thus causing simulation outcomes diverge from real-world observations. Addressing this gap, we propose a novel opinion dynamics model based on hierarchical thinking to describe the opinion evolution on social networks. Individuals are classified into different levels according to their cognitive abilities. They act with bounded rationality at their respective levels to optimize both the promotion of personal opinions and the avoidance of cyberbullying. Through simulation analysis, we found the crucial role of users with high levels of hierarchical thinking. They can discern the opinion climate and articulate their opinion, acting as bridges in the evolution of public opinion. Their opinion can reach the bounded confidence range of more people, thereby enabling polarization to shift to consensus under the same conditions. Furthermore, this effect is independent of individual inherent attributes, which is more in line with real-life scenarios.

Keywords: Deep Learning, Application of Artificial Intelligence, Neural Networks and their Applications
Abstract: Adversarial attacks mislead deep neural network classifiers with slight perturbations, underscoring the necessity for the development of robust defenses to ensure the secure and responsible use of artificial intelligence. Recent research has shown that diffusion-based adversarial purification methods have emerged as a promising defense technique, but often suffer from computational inefficiencies and suboptimal results. To address these issues, we propose RoPAR, an innovative approach that enhances robustness against adversarial attacks by aggregating purified images at intermediate steps of the diffusion process. Our method improves model robustness while reducing the required diffusion steps. We also introduce a technique for reordering Gaussian noise to minimize semantic information loss while removing adversarial perturbations. These enhancements significantly reduce the number of function evaluations from 200 to 6, achieving a robust accuracy of 92.39% against preprocessor-blind PGD attacks on CIFAR-10, a 2.29 percentage point improvement over state-of-the-art. Moreover, our method demonstrates its effectiveness in real-world scenarios, achieving 87.46% accuracy on CIFAR-10C.

Keywords: Biometric Systems and Bioinformatics
Abstract: Fingerprints serve as a primary means of individually identifying individuals. However, employing fingerprints in online mode poses a significant privacy risk, since it is susceptible to several forms of attack. It is plagued by issues related to privacy and security. In response to this, we proposed an innovative approach to convert the original fingerprint into a secure template that may be retained and utilized for authentication purposes. The new templates bear a resemblance to the original human fingerprints and ensure privacy by possessing the characteristic of non-invertibility. This study presented a method for generating highly authentic fingerprint templates that ensure the capacity to revoke and cancel the stolen fingerprint. Throughout the training and testing phase, we utilized the dataset derived from the Vikriti-ID fingerprint. The collection has 25000 distinct fingerprint samples, divided into five classes, with each class containing 5,000 samples. Throughout the testing phase, the comprehensive performance was evaluated based on the matching performance including EER and AUC.

Keywords: Deep Learning, Application of Artificial Intelligence, Neural Networks and their Applications
Abstract: Text-to-speech (TTS) technologies have recently expanded to incorporate natural language prompts for user-friendly control of speech styles, driven by significant advancements in language models. Traditional prompt-based TTS research, however, typically requires large-scale prompt generation that often necessitates costly human annotations. To address this challenge, we propose PromotiCon, a system that leverages prompts generated without human annotations to control emotions in speech. Our model utilizes abundant prompts generated using a large language model. Additionally, we propose an emotion distance-based prompt-speech matching method to appropriately pair the generated prompts with the most resembling speech data. To enhance speaker adaptation, we introduce a semi-supervised approach that allows the joint utilization of multi-speaker data without emotion labels. As a result, our system facilitates zero-shot emotional speech synthesis. Our experimental results confirm the effectiveness of our approach. Audio samples are available at https://promoticon.github.io/.

Keywords: Deep Learning, Application of Artificial Intelligence, Neural Networks and their Applications
Abstract: Adversarial attacks pose a significant threat to security-critical applications by deliberately deceiving model predictions. Numerous works attempt to create robust models by encoding useful information to intermediate representations. However, they still contain too much information about the training data which hinders improving the robustness of the model. To mitigate this issue, we propose a novel approach, CHBaR, that incorporates class-conditioned information into intermediate representations. The class-conditioned information plays the role of weight components which are multiplied with the intermediate representations to produce class-conditioned representations. We utilize an attribution-based explanation method to obtain this class-conditioned information. As a result, the weight components emphasize class-relevant features by highlighting relevant information from the target class. This weighting process easily integrates the target class without complex computations and conceals useless representations, thus enhancing model predictions by masking features unrelated to the class. Extensive experiments demonstrate the effectiveness of our proposed method in enhancing adversarial robustness. Especially, on the SVHN dataset, our proposed method shows an increment of 6.98% points compared to the baseline model in PGD40 adversarial attack with the TRADES training setting.

Keywords: Human-Computer Interaction, Intelligence Interaction, Human-Machine Cooperation and Systems
Abstract: In recent advancements in sensor and artificial intelligence technologies, the reliability of gesture recognition has significantly improved, prompting various industrial fields to adopt this technology. However, most gesture recognition systems rely on optical methods because of their high accuracy, despite require complex and computationally intensive processes. Moreover, these systems are associated with high construction costs and are susceptible to environmental factors. This paper introduces a novel gesture recognition system, which effectively tracks and estimates gestures using cost-effective ultra-wideband (UWB) sensors and inertial measurement units (IMU). The system acquires position data of gesture through UWB sensors and includes essential data processing steps such as the detection and removal of abnormal data via IMU, data smoothing with a Kalman filter, and data normalization and scaling. Notably, normalization and scaling are achieved by converting the position data into grayscale images, ensuring the consistency of data features and enhancing gesture recognition accuracy across diverse users. The proposed system employs a convolutional neural network (CNN) model to estimate gestures from these images. Comparative analyses demonstrate that the proposed system exhibits superior gesture classification performance compared to systems utilizing a long-short term memory (LSTM) model and those employing the same CNN model without the aforementioned data processing steps. Therefore, this system is not only cost-effective but also efficiently tracks and estimates gestures, offering significant improvements over existing methods.

Keywords: Human-Machine Interaction, Intelligence Interaction
Abstract: As the techniques of autonomous robots advance, there is an increasing demand for robots to provide explanations for their behavior. There are two commonly used explanation types. The first type emphasizes that a robot’s policy is the best (or only) option that satisfies a specific property produced by its decision-making algorithms. The second explanation type is used when a robot fails and describes the cause of an error state that led to the failure. This paper proposes a new explanation type derived from a robot's proficiency self-assessment. The proposed explanation type not only supplements the first explanation type under typical operating conditions but also includes the second explanation type when the robot fails. The proposed explanation type is based on assumption-alignment tracking (AAT), a novel method for robot proficiency self-assessment. AAT provides three pieces of information for explanation generation: (1) assessment of assumptions veracity on which the robot's generators rely; (2) proficiency assessment measured by the probability that the robot will successfully accomplish its task; (3) counterfactual proficiency assessment computed by hypothetically varying assumptions. The information provided by AAT fits the situation awareness-based framework for explainable artificial intelligence. Examples of generated explanations are demonstrated using a simulated robot setting up a table with different blocks.

Keywords: Human-centered Learning, Human-Machine Interaction, Human-Machine Cooperation and Systems
Abstract: In recent years, robots and autonomous systems have become increasingly integral to our daily lives, offering solutions to complex problems across various domains. Their application in search and rescue (SAR) operations, however, presents unique challenges. Comprehensively exploring the disaster-stricken area is often infeasible due to the vastness of the terrain, transformed environment, and the time constraints involved. Traditional robotic systems typically operate on predefined search patterns and lack the ability to incorporate and exploit ground truths provided by human stakeholders, which can be the key to speeding up the learning process and enhancing triage. Addressing this gap, we introduce a system that integrates social interaction via large language models (LLMs) with a hierarchical reinforcement learning (HRL) framework. The proposed system is designed to translate verbal inputs from human stakeholders into actionable RL insights and adjust its search strategy. By leveraging human-provided information through LLMs and structuring task execution through HRL, our approach not only bridges the gap between autonomous capabilities and human intelligence but also significantly improves the agent's learning efficiency and decision-making process in environments characterised by long horizons and sparse rewards.

Keywords: Human-Machine Cooperation and Systems, Assistive Technology, Human-Machine Interaction
Abstract: Work-related musculoskeletal disorders (WMSD) continue to be a significant cause of work incapacity. Exoskeletons have the potential to prevent these disorders, with passive exoskeletons already proving their usefulness but also displaying limitations. However, there is currently a lack of suitable methods to control active exoskeletons, which offer additional advantages like supporting the user just when needed. This paper proposes a sensor system and a method to classify different activities based on the kinematic and activation signals of the used power tools. First, requirements and thresholds for the sensor system and its signals were derived from representative activities in the construction environment. It then introduces a sensor system that collects the tool's kinematic signals and activation. The sensor system consists of an inertial measurement unit (IMU), pressure sensor, and WiFi-capable micro-controller to stream the data. Based on the signals, a threshold-based algorithm, capable of identifying six predefined activities, is presented. The paper presents a suitable test course based a real situation in the construction industry to evaluate the proposed sensor system and algorithm. With the developed test course a study is conducted and the activities are classified based on the signals of the sensor system. The results demonstrate that the defined activities can be distinguished based on the kinematic and activation signals of the power tools.

Keywords: Application of Artificial Intelligence, Cloud, IoT, and Robotics Integration, AIoT
Abstract: CiRA CORE is a central hub designed to connect AI technology creation with practical application, making it easier to work with ROS (Robot Operating System) and link different systems through a user-friendly drag-and-drop interface. This approach removes the need for extensive coding, making the platform accessible to those with minimal programming experience. CiRA CORE offers a comprehensive suite of features for AI development and robot control, including algorithm creation, AI model training, and device integration commonly used in industrial settings. It supports tasks like image recognition and facilitates data storage, labeling, and integration with other systems for data-driven AI development. Overall, CiRA CORE aims to democratize AI development and robot control, simplifying AI development for Industry 4.0 applications, and leading to increased efficiency, reduced costs, and improved safety in industrial processes. This paper reports the progress of the CiRA CORE training modules funded by the SMCS TEAM Program Award. The project has completed the design of a 6-axis robot 3D training kit and simulation models for CiRA CORE training modules. The next steps involve developing 3D-printed robots and training materials. The main goal is to democratize advanced robotics and AI by simplifying integration through a visual, node-based programming interface. This approach reduces the need for complex coding, making these technologies accessible to users with limited programming experience. This initiative aims to foster widespread adoption in business and industrial settings, aligning with IEEE SMC's mission to promote professional growth and innovation in robotics and AI.

Keywords: Consumer and Industrial Applications, Large-Scale System of Systems, Manufacturing Automation and Systems
Abstract: Elon Musk's successful "move fast and break things" strategy is based on the fact that in many cases, we do not need to satisfy all usual constraints to be successful. By sequentially trying smaller number of constraints, he finds the smallest number of constraints that are still needed to succeed -- and using this smaller number of constrains leads to a much cheaper (and thus, more practical) design. In this strategy, Musk relies on his intuition -- which, as all intuitions, sometimes works and sometimes doesn't. To replace this intuition, we propose an algorithm that minimizes the worst-case cost of finding the smallest number of constraints.

Keywords: AI and Applications, Application of Artificial Intelligence, Computational Intelligence
Abstract: Currently, population in Japan has been aging at a speed unparalleled in other countries, and countermeasures against aging population and the worsening of disease for people with disabilities have become urgent issues. Pneumonia and aspiration pneumonia are the leading causes of death. It is said that swallowing function tends to decline from around age of 40, however it is important to keep it in good condition without deteriorating function as much as possible. The gold standard for swallowing functional evaluation is swallowing contrast testing, however X-ray exposure disables to repeat testing. In addition, Repetitive Saliva Swallowing Test (RSST) of screening test is difficult for self-check. Therefore, in this study, we develop a system to self-evaluate swallowing ability for keeping swallowing function healthy. It is proposed by applying optical flow and artificial intelligence of DeepLabcut. As a result, we were able to visualize movement of the larynx during swallowing.

Keywords: AIoT, Computational Intelligence, Soft Computing, Socio-Economic Cybernetics
Abstract: Agriculture is one of the sources of income a region can rely on to support its economy. Traditional agriculture relies primarily on human performance and observation, resulting in greater production costs and, subsequently, higher selling prices. Artificial intelligence-based technology can be used to reduce production costs, increase productivity, and provide consumer convenience. An indicator that is easy to interpret in measuring the quality and optimization of plant growth is the visualization of the condition of the leaves. The artificial intelligence technique that can be implemented in this regard is the object detection model. However, the challenge is the complex, multi-object, and multi-intersection condition of the leaves, which causes the model to be less optimal in conducting classification and detection tasks regarding whether the leaf condition is good or not. A YOLOv7 model will be employed in order to detect leaf quality, whether in an “optimal” or “not optimal” condition. To enhance the model's performance by improving accuracy through feature extraction enhancement, YOLOv7 will be integrated with the attention module, called the convolutional block attention module (CBAM). The case study in this research is detecting a mango plant which is one of the plants that can provide a high economic impact and the object observed is the mango plant leaf. Several previous studies related to the implementation of attention modules in object detection include the improved pest-YOLO for real-time pest detection by combining YOLOv3 with efficient channel attention (ECA) and a transformer encoder. The ECA module and transformer encoder were integrated into the backbone and neck block systems of YOLO [1]. The lightweight YOLO model combined with SE-CSPGhostnet by improving the backbone block which employs squeeze-and-excitation networks (SENet) and a convolution technique consisting of regular convolution and ghost convolution [2]. There is a highlighted improvement of YOLOv7 compared to the previous version of YOLO, which is Extended Efficient Layer Aggregation Networks (E-ELAN). YOLOv7's learning ability is enhanced by using this network while maintaining the transition layer's architecture. E-ELAN enhance

Keywords: Assistive Technology, User Interface Design, Companion Technology
Abstract: Web forms play a pivotal role in digital interfaces but frequently pose significant accessibility challenges. This paper explores the main barriers to creating accessible web forms and investigates how generative AI technologies can provide solutions. We highlight core issues such as accurate labeling, keyboard navigation, error management, focus control, visual design factors, placeholder text usage, assistive technology compatibility, handling of complex inputs, responsive design, cognitive load reduction, and ongoing testing. For each of these challenges, we assess its effect on accessibility and present innovative AI-driven strategies. Our findings illustrate how AI can streamline the development process by automating label generation, improving tab indexing, enhancing real-time error detection, refining focus control, offering contrast improvement suggestions, and simulating interactions with assistive technologies. We conclude that incorporating generative AI into web form development can markedly improve accessibility, making digital experiences more inclusive for users of all abilities. This not only supports compliance with legal and ethical standards but also fosters a more inclusive online environment, enhancing user satisfaction and overall experience.

Keywords: Application of Artificial Intelligence, Swarm Intelligence, AI and Applications
Abstract: The production of palm oil on a commercial scale is labour intensive with many of its processes handled by humans. In some countries, there can be as many as 500,000 plantation workers in the palm oil sector involved in labour intensive work in large plantations. However, such dependence on humans for low skill manual work has led to many problems. Unmanned aerial vehicles (UAVs) have been seen as a possible alternative to support some of these processes that require low skills in the palm oil industry. However, the flying time of the UAVs is finite and hence it is important to maximize the number of palm trees that it can service. In this paper, an Ant Colony System (ACS) with a novel path constructor was used to identify good flight paths for UAVs in large palm oil plantations to help improve the efficiency for some of the agricultural activities. Good results were obtained for various data sets especially when compared with the standard ACS as well as those by the human experts.

Keywords: Computational Intelligence, AI and Applications, Machine Learning
Abstract: The Broad Learning System (BLS) has been established as an effective flat network alternative to Deep Neural Networks (DNNs), delivering high efficiency while achieving competitive accuracy. Despite its advantages, the incremental learning methods of BLS face challenges in stability and computation when expanding with new nodes or input. We introduce two novel incremental learning algorithms based on factorization updates for BLS that optimize node and input additions to overcome these limitations. Our node addition algorithm utilizes QR decomposition and Cholesky factorization, using the update of the Cholesky factor instead of pseudo-inverse computations. For input addition, we propose an iterative Cholesky factor update algorithm. Our algorithms demonstrate not only faster computation compared to the existing BLS but also improved testing accuracy on the MNIST or Fashion-MNIST dataset. This work presents a significant step forward in the practical application and scalability of BLS in various data-dense environments.

Keywords: Deep Learning, AI and Applications
Abstract: In recent years, object detection models DETRs based on Transformer architecture have played a huge role in various fields. However, the DETR series models are not satisfactory in small object detection. Mainly due to the huge amount of calculation of DETR, a lot of feature information will be lost in the feature fusion stage and the low tolerance of small objects to Intersection over Union (IoU). In order to solve the above problems, we propose a near real-time detection model RTS-DETR. In this paper, we revisit RT-DETR, which effectively handles multi-scale features by decoupling intra-scale interactions and cross-scale fusion, but this will lose a lot of positive local information. To this end, we have improved the efficient hybrid encoder. We propose a new positional encoding method that enables the hybrid encoder to more accurately convert the input feature sequence into a high-dimensional representation, and propose a new feature fusion module to enhance the model's ability to capture local features. Furthermore, in order to improve the tolerance of small objects to IoU, we combine Normalized Wasserstein Distance (NWD) with Shape-IoU for the optimization model. This method more accurately takes into account the shape and size of objects, thereby improving detection accuracy. Our model achieves an accuracy of 38.8% (in terms of mAP_{@0.5}) on the widely used VisDrone dataset, which improves the accuracy by 2.5% compared to RT-DETR with ResNet-18 as the backbone network.

Keywords: Application of Artificial Intelligence, Deep Learning, Knowledge Acquisition
Abstract: Large language models (LLMs), such as ChatGPT, have demonstrated remarkable capability in question answering but face challenges when it comes to knowledge-based reasoning, such as limited training data and hallucination. To address these challenges, integrating LLMs with knowledge graphs (KGs) has emerged as a promising solution. However, the cost associated with training and inference of LLMs is high. Our method integrates the Retrieval-Augmented Generation (RAG) paradigm, incorporating relevant information from KGs alongside the question to enhance LLMs' reasoning process without training. Moreover, we propose a novel concept of self-knowledge motivation to reduce the overhead of inference, which prompts LLMs to integrate retrieved information with their internal knowledge for reasoning before seeking additional queries to KGs. Experimental results showcase improvements in answer accuracy and a reduction in LLMs' API calls compared to the latest published state-of-the-art (SOTA) method employing an identical paradigm, underscoring the efficiency and effectiveness of our method.

Keywords: Application of Artificial Intelligence, Computational Intelligence, Neural Networks and their Applications
Abstract: Building an autonomous intelligent agent capable of carrying out web automation tasks from descriptions in natural language offers a wide range of applications, including software testing, virtual assistants, and task automation in general. However, recent studies addressing this problem often require manually constructing of prior human demonstrations. In this paper, we approach the problem by leveraging the idea of reinforcement learning (RL) with the two-phase mechanism to form an agent using LLMs for automating computer tasks without relying on human demonstrations. We evaluate our LLM-based agent using the MiniWob++ dataset of web-based application tasks, showing that our approach achieves 85% success rate without prior demonstrations. The results also demonstrate the agent's capability of self-improvement through training.

Keywords: Artificial Social Intelligence, AI and Applications, Machine Learning
Abstract: With few-shot learning abilities, pre-trained language models (PLMs) have achieved remarkable success in classification tasks. However, recent studies have shown that the performance of PLM is vulnerable due to different prompts and the instability of the prompt-based learning process. To address this challenge, we explore appropriate perturbation addition of adversarial training and integrate the global knowledge of the full-parameter fine-tuned pre-trained language model(PLM). Specifically, we propose an adversarial prompt learning model (ATPET) and ATPET with fine-tuning(ATPETFT), incorporating ATPET with fine-tuning knowledge into the prompt learning process. Through extensive experiments on several few-shot classification tasks and challenging data settings, we demonstrate that our methods consistently improve the robustness while maintaining the effectiveness of PLMs.

Keywords: Image Processing and Pattern Recognition, AI and Applications, Neural Networks and their Applications
Abstract: In focusing tasks on moving targets, traditional methods that rely on maximizing contrast struggle to capture moving objects due to insufficient focusing speed. Deep learning-based methods have attempted to directly predict the optimal focal length for the target; however, due to low prediction accuracy, they often lead to out-of-focus situations when capturing moving objects. In recent years, some approaches have utilized reinforcement learning to automatically explore focal length adjustment patterns, thus achieving better results than traditional methods. However, these approaches have not considered the motion characteristics of the targets, leading to a need for further improvement in focusing performance. To overcome these limitations, we introduce a motion-based feature and deep reinforcement learning-driven autofocus algorithm named MF-DRLAF for moving targets. This novel method tracks the object, predicts its motion state through feature extraction, and uses deep reinforcement learning to dynamically adjust the focus. We utilize a self-attention mechanism to adaptively learn various motion patterns and employ a feature pool structure to enhance processing efficiency. Experiments and real-world testing on a Google Pixel3 demonstrate that our approach significantly enhances autofocus performance on moving objects, highlighting its potential for broader imaging applications. This approach offers a promising direction for future development in autofocus technology.

Keywords: System Modeling and Control, Intelligent Power Grid, Control of Uncertain Systems
Abstract: Load frequency control (LFC) of a two-area connected electric power system is vital for maintaining grid stability and reliability by matching power generation with load demand. Thus, this work proposes an analytical approach for designing a fractional order (FO) controller to regulate the LFC of a two-area connected electrical power system with time delay. First, the interconnected electrical power system is accurately modelled as a FO system with time delay. Then, the FO controller is designed using the internal model control (IMC) technique, where a low-pass filter (LPF) is considered to mitigate the effect of the disturbances. The tuning parameter of the designed FO involves a single tuning parameter, which is analytically designed using gain crossover frequency criteria. The disturbance and parametric uncertainty analyses have been carried out to analyze the efficacy of the proposed method under the variation of tuning parameter. Then, the frequency and tie-line power fluctuations are estimated under nominal and parametric uncertainty conditions. Also, its performance has been compared to recent state-of-the-art techniques for precise efficacy analysis.

Keywords: System Modeling and Control, Cyber-physical systems, Modeling of Autonomous Systems
Abstract: Synchronous language is routinely used to model safety-critical control systems. In recent years, it is gradually being applied to cyber-physical systems (CPS) which emphasise high levels of correctness and safety. It is based on the assumption that the system reacts instantaneously to input events and can compute the output before the next input event, so it is well suited for expressing temporal logic. However, it lacks effective constructs for expressing spatial properties in CPS. Moreover, spatio-temporal properties in CPS are indispensable, requiring not only qualitative analysis but also quantitative analysis. Therefore, we propose SELus, a new synchronous language based on Lustre, to provide the capability of modeling spatio-temporal properties in CPS, enabling the representation of spatial topological relationships and the performance of quantitative analysis on them. To formally verify the SELus model, we introduce a set of mapping rules to transform the SELus model into the Ptolemy II model. The resulting Ptolemy II model is used in Ptolemy II to perform quantitative analysis of the SELus model. Experiments are conducted on lane changing system, showcasing the usability and effectiveness of our language.

Keywords: Control of Uncertain Systems, Modeling of Autonomous Systems, Robotic Systems
Abstract: In this paper, we investigate the motion of wheeled mobile robots on rough terrains modeled as noisy nonholonomic constraints. Such constraints are the natural extension of ideal nonholonomic constraints when the Stratonovich process is directly introduced in the constraint equations. The resulting stochastic model can capture motion on rough surfaces, random deformation in the wheel-ground contact, or stochastic loss/gain of traction. We study a differential robot with ideal noisy and affine noisy constraints, where each case models a certain aspect of motion on rough terrains. We then investigate their corresponding stochastic dynamics and the propagation of mean and covariance through Monte-Carlo simulations. The proposed model for roving rough terrains has the potential to serve as the stochastic model employed in model-based motion planning, pose estimation, and control of rover systems. The main challenge will be dealing with the nonlinear appearance of the noise and its feedback in the equations of motion.

Keywords: Autonomous Vehicle, System Modeling and Control, Modeling of Autonomous Systems
Abstract: To tackle the twin challenges of limited battery life and lengthy charging durations in electric vehicles (EVs), this paper introduces an Energy-efficient Hybrid Model Predictive Planner (EHMPP), which employs an energy-saving optimization strategy. EHMPP focuses on refining the design of the motion planner to be seamlessly integrated with the existing automatic driving algorithms, without additional hardware.It has been validated through simulation experiments on the Prescan, CarSim, and Matlab platforms, demonstrating that it can increase passive recovery energy by 11.74% and effectively track motor speed and acceleration at optimal power. To sum up, EHMPP not only aids in trajectory planning but also significantly boosts energy efficiency in autonomous EVs.

Keywords: Autonomous Vehicle, Modeling of Autonomous Systems, Intelligent Transportation Systems
Abstract: The sensor configuration of an autonomous vehicle (AV) is determined in the early development phase when specific perception algorithms are not yet available. Therefore, approaches based on synthetic raw data are necessary to evaluate different configurations. One sensor type used in AV is LiDAR, but developers should carefully consider the amount and placement of the sensors due to their high costs. In this contribution, we propose the Omni-Lidar Evaluation Score (OLES), a novel metric to evaluate different LiDAR configurations based on their simulated raw data. Our OLES metric combines information-theoretic quantities with coverage-based metrics, considering both the spatial coverage and the uniformity of a LiDAR point cloud distribution. We show the need for a new metric and provide details on implementing OLES using the open-source simulator textit{CARLA}. We demonstrate the effectiveness of our new metric in a simulation study and highlight its usefulness in the early phases of vehicle development. This research provides a means to evaluate the quality of LiDAR configurations and provides a basis for further optimizing sensor setups for AVs.

Keywords: Consumer and Industrial Applications
Abstract: Considering two eco-label strategies, self-label and certification-label, we construct a duopoly competition model encompasses both green product and ordinary product manufacturing enterprises. By Investigating the optimal eco-label standards, we explore the product pricing, and profits for enterprises facing green-sensitive consumers and price-sensitive consumers. The we analyze the optimal eco-label selection for green enterprises in different preference markets. Research indicates that the green quality standards and product prices under certification labels are invariably higher than those under self-label. However, the choice of eco-label by enterprises is influenced by consumers' individual intrinsic preferences; in price-sensitive markets, enterprises tend to adopt self-label; In green-sensitive markets, when the value of consumers' individual intrinsic preferences is below a certain threshold, enterprises will prioritize certification labels. Additionally, the profits of enterprises in green-sensitive markets are generally higher than those in price-sensitive markets, enterprises should highlight the advantages of green quality and guide consumers to prefer green attributes more when formulating promotional strategies.

Keywords: Consumer and Industrial Applications, System Architecture
Abstract: With the proliferation of artificial intelligence, Computer-Aided Design (CAD) is being transformed into Computer-Automated Design (CAutoD). In this paper, the advent of Large Language Models (LLMs) introduces new opportunities for CAutoD. This study develops AutoForma, an LLM-based multi-agent system, for automatic conversion from natural language descriptions to 3D models. By harnessing the comprehension capabilities of LLMs, AutoForma streamlines the CAutoD workflow by efficiently translating design intents into precise models in CAD. Through a comprehensive set of evaluations, AutoForma is seen to offer automation performance across various design tasks, particularly in generating non-standard parts that meet specific requirements, with higher efficiency and accuracy than using just an LLM like GPT-4.

Keywords: System Modeling and Control, Electric Vehicles and Electric Vehicle Supply Equipment, Autonomous Vehicle
Abstract: Advanced control and safety systems are crucial for electric vehicles, and the accurate estimation of the tire-road friction coefficient (TRFC) is crucial for developing effective safety control strategies. The hybrid data-mechanism model (HDMM), introduced in this paper, addresses the performance challenges posed by the inaccuracies of physical models and the limited interpretability of data-driven models in tire force estimation for TRFC estimation.Tire dynamics often exhibit transient responses, while mechanism-based models(MBM) typically reflect steady-state characteristics. Neglecting transient characteristics leads to a decrease in model accuracy.A neural network is used to learn the transient response characteristics of tire dynamics.These characteristics are then integrated with the steady-state tire forces from MBM to estimate the lateral and vertical forces acting on the wheel.The estimated tire forces serve as virtual measurements to calibrate parameters in the TRFC estimator, based on the Unscented Kalman Filter (UKF). During real-world vehicle tests, the proposed method reduced the Mean Error (ME) in lateral and vertical forces by 1271.85 N and 996.7 N, respectively, compared to the estimated tire forces from MBM. Additionally, the estimated TRFC converged to the reference value approximately 40ms earlier than the result from the MBM, with an estimated deviation within 0.1.

Keywords: Intelligent Transportation Systems, Distributed Intelligent Systems, Smart Buildings, Smart Cities and Infrastructures
Abstract: Traffic flow prediction assumes a pivotal role in aiding governments and companies accurately forecast changes in vehicle volume, consequently enhancing transportation efficiency and facilitating vehicle travel. Presently, the majority of traffic flow prediction methods rely on centralized learning strategies, which entail the transmission of substantial data and may jeopardize user privacy. To address this issue, we propose a Federated Learning-based Attention Graph Convolutional Network (FLSTAGCN) algorithm for traffic flow prediction. Firstly, we develop a Spatial-Temporal Attention Graph Convolutional Network (STAGCN) method that employs attention mechanism to proficiently extract spatial-temporal features from traffic flow data, augmenting the model's learning capabilities. Subsequently, within the aggregation mechanism of Federated learning, we devise a bespoke optimal selection to enhance training accuracy and reduce communication costs in traffic flow prediction scenarios. Finally, we integrate Federated Learning with STAGCN and utilize the optimal selection protocol to designate participants for transmitting optimal parameters. The Experimental results substantiate that our approach outperforms advanced deep learning approaches in terms of traffic flow prediction performance while ensuring the privacy and security of traffic data.

Keywords: Autonomous Vehicle, Decision Support Systems, Adaptive Systems
Abstract: Autonomous driving demands sophisticated control systems that optimize safety, performance, passenger comfort, and fuel efficiency. This study proposes a steering control system that integrates the Deep Deterministic Policy Gradient (DDPG) for speed planning with a novel feedback mechanism based on Subjective Vertical Conflict (SVC) in the reward function. Using simulations in MATLAB and Simulink, we evaluate the system's performance across various thresholds of SVC, examining its impact on ride comfort, fuel efficiency, and vehicle behavior during lane changes. Results reveal a trade-off relationship between ride comfort and fuel efficiency, with lower SVC thresholds generally improving comfort but potentially increasing steering input. Additionally, excessively low SVC thresholds degrade target-reaching performance and lengthen lane change distances, highlighting the need for careful parameter tuning. Overall, our findings demonstrate the potential of reinforcement learning-based steering control systems to optimize multiple evaluation criteria simultaneously while emphasizing the importance of balancing trade-offs in autonomous driving scenarios.

Keywords: Autonomous Vehicle, Control of Uncertain Systems, System Modeling and Control
Abstract: This paper presents a novel robust path-tracking controller for autonomous ground vehicles. Environmental and vehicle factors like variation in road conditions and varying speed can adversely affect autonomous ground vehicles' path-tracking capability. A polytopic linear parameter varying model for autonomous ground vehicle that accounts for system uncertainties with varying speeds and road conditions is formulated. Then, an H_∞ based robust path-tracking controller is developed using this model to minimise the vehicle's lateral velocity, heading error, and slip angle. Simulation results comparing the proposed controller with a conventional robust controller are presented. The findings show that the proposed controller performs well and is more effective than the conventional robust controller.

Keywords: Autonomous Vehicle, Modeling of Autonomous Systems, Cooperative Systems and Control
Abstract: Advanced driver assistance system (ADAS) is effectively promoting the vehicular automation level and it is critical to ensure its functional safety. While existing analysis mainly focuses on individual applications of ADAS, safety violations in the overall system can be found by extensive road tests, which are not only costly in terms of time and money but also lack a formal safety guarantee. This is because tests may not cover all driving scenarios, especially the ones that involve discrete mode switching. In this paper, we focus on the longitudinal vehicle motion and provide a pipeline to perform safety verification for all the related ADAS applications. To that end, we specify safety constraints and boundaries for a vehicle’s longitudinal cruising and collision avoidance and validate a longitudinal dynamic model against the high-fidelity simulation software CarSim. Then we define hybrid automata to describe the closed-loop system composed of the vehicle dynamics and the ADAS. Finally, by computing the reachable sets of the hybrid automata and comparing them with the specified safety boundaries, the ADAS is verified. Numerical experiments demonstrate the efficacy of the proposed approach.

Keywords: Intelligent Transportation Systems
Abstract: Accurate traffic Flow Prediction can assist in traffic management, route planning, and congestion mitigation, which holds significant importance in enhancing the efficiency and reliability of intelligent transportation systems (ITS). However, existing traffic flow prediction models suffer from limitations in capturing the complex spatial-temporal dependencies within traffic networks. In order to address this issue, this study proposes a multi-segment fusion-enhanced spatial-temporal graph convolutional network (MS-STGCN) for traffic flow prediction with the following three-fold ideas: a) building a unified spatial-temporal graph convolutional framework based on Tensor M-product, which capture the spatial-temporal patterns simultaneously; b) incorporating hourly, daily, and weekly components to model multi temporal properties of traffic flows, respectively; c) fusing the outputs of the three components by attention mechanism to obtain the final traffic flow prediction results. The results of experiments conducted on two traffic flow datasets demonstrate that the proposed MS-STGCN outperforms the state-of-the-art models.

Keywords: Evolutionary Computation, Computational Intelligence, Application of Artificial Intelligence
Abstract: In this research, we evolve gaits for an arachnid-inspired robot. The method used is an expansion upon previous research on the incremental evolution of gaits for hexapod robots with two degrees of freedom per leg, which we now apply to a more complex, eight-legged robot with three degrees of freedom per leg. Incremental evolution handles gait generation for legged robots in two discrete increments. The first increment uses a cyclic genetic algorithm to learn the activations (pulse instructions to the servos) required for each leg to perform a single-leg cycle. This learning program takes into account the way each leg is mounted on the body and the range of movement provided by the three servos on each leg to produce a smooth, straight, and efficient leg cycle. The second increment uses a genetic algorithm to select the best combination of leg cycles for each leg and to learn the timing to execute each leg cycle to coordinate them all together into a single gait. In this work, we learn the gait incrementally in a simulation and transfer the final gaits to the real robot to confirm the method’s viability.

Keywords: Swarm Intelligence, Evolutionary Computation, Computational Intelligence
Abstract: Leading exemplars play significant roles in updating particles to seek optimal solutions for Particle Swarm Optimization (PSO). Along this road, this paper devises an Individual-level Dominant Exemplar Selection (IDES) framework for PSO, giving rise to a new PSO variant named IDESPSO. Specifically, instead of using their own personally best positions and the globally best position of the entire swarm to update particles, IDES first randomly chooses two different exemplars for each particle from all personally best positions. Then, it compares the two selected exemplars with the personally best position of this particle. Based on the comparison results, different updating strategies are utilized to update different particles. This method notably enriches the variety among the chosen leading exemplars, thereby substantially bolstering the updating diversity of particles. Under IDES, this paper further develops seven selection strategies to help IDESPSO pick up promising exemplars for particles to evolve. Specifically, the seven selection schemes are the roulette wheel selection, the tournament selection, and five hybridizations of two basic models. A series of experiments have been undertaken on the universally used CEC2014 problem suite to compare IDESPSO with the seven selection schemes and two classic PSOs. The empirical results show that IDESPSO paired with anyone of the seven selection methods, markedly outperforms the two classical PSO variants, highlighting its significant performance.

Keywords: Evolutionary Computation, Computational Intelligence, Machine Learning
Abstract: Traffic signal control (TSC) problems have re- ceived increasing attention with the development of the smart city. Reinforcement learning (RL) models TSC as a Markov decision process and learns the timing relationship of traffic scheduling from massive historical data. Due to the uncertainty and mutability of TSC problems, existing RL methods face bottlenecks in diversity and are easy to be trapped into local optima. To alleviate this predicament, this paper combines evolutionary optimization and RL to propose an evolution- ary algorithm-assisted reinforcement learning (EARL-Light) method for TSC problems. EARL-Light is a population-based algorithm, in which one individual represents a policy and a population of individuals are evolved to search for near-optimal policies. The diversified search ability of evolutionary optimiza- tion can help the algorithm get rid of local optima for global optimization and the rapid learning based on the gradient of RL can achieve fast convergence. Extensive experiments on seven real-world traffic datasets demonstrates that EARL-Light achieves shorter travel time with fast convergence.

Keywords: Evolutionary Computation, Computational Intelligence, Application of Artificial Intelligence
Abstract: Target tracking has broad applications like disaster relief, and unmanned aerial vehicles (UAVs) have been universally applied in target tracking in recent years. Due to the strong responsiveness to deceptive reward signals and diverse exploration, evolutionary reinforcement learning (ERL) is a more noteworthy option for training UAVs than common reinforcement learning. However, for ERL contains too many agents, its training efficiency is not satisfactory enough. To address this shortcoming, this paper proposes an evolutionary reinforcement learning with double replay buffers (ERLDRB) for UAV online target tracking problem. Firstly, considering the energy consumption and the possible delay of feedback signals to the UAV, a more realistic model of UAV online target tracking problem is designed. Then based on the problem formulation, ERLDRB utilizes a double experience replay buffers technique to increase learning efficiency in the training stage, which can better solve real-world UAV online target tracking problem. Simulation results show that ERLDRB outperforms multiple contrasting algorithms on the designed model.

Keywords: Evolutionary Computation, Swarm Intelligence, Computational Intelligence
Abstract: Ant colony system algorithm (ACS), as an important evolutionary computation (EC) algorithm, has demonstrated significant advantages in solving complex optimization problems. However, traditional EC algorithms and traditional ACS algorithm often face the challenge of slow computational speed when dealing with large-scale problems. In recent years, matrix-based EC approaches have been proposed to accelerate the computational speed, which has obtained promising results in dealing with large-scale problems. However, most existing matrix-based EC algorithms are designed for continuous optimization problems, while the matrix-based approach integrated with ACS has not attracted enough attention, which will be efficient for solving large-scale discrete optimization problems. Therefore, in this paper, we propose a matrix-based ACS (MACS) algorithm and apply it to solve the traveling salesman problem (TSP). MACS is an innovative improvement over the traditional ACS algorithm, utilizing matrix operations to parallelly let ants select city and update pheromone. Experimental results show that the MACS algorithm has significantly better efficiency in accelerating computational speed while maintaining the remarkable problem-solving ability in solving large-scale TSP.

Keywords: Fuzzy Systems and their applications, Computational Intelligence
Abstract: On the one hand, to model experts' preferences in group decision-making, intuitionistic reciprocal preference relations have widely been used because they allow for accommodating hesitation degrees, which are inherent to all decision-making processes. On the other hand, an optimization of information granularity distribution has recently been applied to establish consensus during group decision-making processes. Concretely, a symmetric and uniform distribution of information granularity has been considered for intuitionistic reciprocal preference relations. However, there exist other protocols of information granularity distribution that could be used. Therefore, we aim to analyze all the information granularity distribution protocols and determine their effectiveness in building consensus through intuitionistic reciprocal preference relations. The performance of the different protocols is discussed by conducting some numerical experiments that help provide insights into the effectiveness of the protocols to build consensus.

Keywords: Neural Networks and their Applications, Machine Learning, AI and Applications
Abstract: Branch prediction stands as a key bottleneck in enhancing CPU performance, particularly evidenced by an average of around 10 mispredicted hard-to-predict(H2P) branches per benchmark in SPEC 2017 by current neural network methods. To improve, this paper proposed a Clustered Multitask Learning and Branch Attention Mechanism-Based Branch Predictor (CMA-BP). Clustered multi-task learning enhances model generalization, and branch attention extracts preferences of different branches for global history. Thus, CMA-BP efficiently aggregates branches with similar features, reducing training complexity. Experimental results show that CMABP outperforms existing predictors in accuracy significantly and in the number of parameters required. By advancing the state-of-the-art in branch prediction, this work has important implications for future high-performance computer architecture design

Keywords: Deep Learning, Machine Learning, Neural Networks and their Applications
Abstract: High-resolution remote sensing image object detection is an important research area in remote sensing information processing and has substantial practical applications. This domain presents unique challenges, including variable object scales, complex backgrounds, prevalent small objects, and densely arranged items, distinguishing it from traditional object detection in natural images. This paper proposes a novel object detection algorithm(RS-DETR), which builds upon the DETR framework and integrates the Swin Transformer. The algorithm features a dual-branch structure in its feature extraction module, markedly improving detection accuracy, especially for objects of varying scales. The addition of the GAM convolutional attention mechanism allows the model to concentrate more effectively on relevant regions, minimizing background complexities. Moreover, we have included the scale-invariant intersection over union (SIoU) loss function to enhance the precise localization of closely packed objects. To demonstrate the efficacy of the algorithm, RS-DETR was applied to the HRSC2016 and NWPU VHR-10 datasets. The results show average detection accuracies of 86.1% and 57.9% on these datasets, respectively, outperforming the baseline models by 1.1% and 0.9%, respectively.

Keywords: Deep Learning, Neural Networks and their Applications, Machine Learning
Abstract: Captioning Whole Slide Images (WSIs) for pathological analysis is an essential but not extensively explored aspect of computer-aided pathological diagnosis. Challenges arise from insufficient datasets and the effectiveness of model training. Generating automatic caption reports for various gastric adenocarcinoma images is another challenge. In this paper, we introduce a hybrid method referred to as TransUAAECapGen to generate histopathological captions from WSI patches. The TransUAAE-CapGen architecture consists of a hybrid UNet-based Advereasrial Autoencoder (AAE) for feature extraction and a transformer for caption generation. The hybrid UNet-based AAE extracted complex tissue properties from histopathological patches, transforming them into lowdimensional embeddings. The embeddings are then fed into the transformer to generate concise captions. Our proposed method is validated using the PatchGastricADC22 dataset. The TransUAAE-CapGen model provides the best estimated accuracy of BLEU-4 = 86.8%, METEOR = 59.6%, a ROUGE = 89.3%, and CIDEr = 7.72%. Experimental analysis indicates that the TransUAAE-CapGen architecture outperforms the traditional LSTM-based model for the caption generation task. Our findings reveal that the proposed architecture can effectively generate accurate and precise reports for medical image analysis.

Keywords: Deep Learning, Machine Vision, Image Processing and Pattern Recognition
Abstract: Learning-based image compression techniques combined with current Transformer models and with checkerboard context models have shown the excellent Rate-Distortion performance. However, the mixed structure still has room for optimization in terms of redundancy information and decoding efficiency, while the checkerboard context model has redundancy in capturing correlations between latent representations. To solve these problems, we propose an innovative framework that combines a mixed Transformer-CNN structure with a checkerboard context model. Specifically, we introduce a ``Checkerboard Channel-wise Entropy Module" to improve coding efficiency of utilizing contexts through a two-channel decoding method with checkerboard contexts. Then, we propose the ``In-slice Odd-even Context", which improves the handling of spatial redundancy information by adding additional spatial contexts by introducing a checkerboard context model to the original mixed structure with channel contexts and global contexts. Extensive experimental results demonstrate that our proposed method outperforms JPEG, BPG and previous learned image compression on the Kodak dataset.

Keywords: Machine Learning, Neural Networks and their Applications
Abstract: The Broad Learning System (BLS) features a simple yet efficient network structure, with its core being the fast and random generation of hidden layers; however, this generation method not only fails to effectively capture the nonlinear characteristics in the task, but also generates certain 'redundant nodes', which can negatively affect its learning capabilities. In this study, we propose an improved version of BLS, named the KEFBLS, aimed at enhancing the feature extraction capability of the hidden layer through the integration of multi-kernel technology and network sparsification strategies, complemented by deeper feature extraction using random Fourier features. the KEFBLS first combines polynomial and wavelet kernels to boost the nonlinear mapping capabilities of data; then, it applies the elastic-net method to refine the BLS objective function, removing low-impact hidden layer nodes to reduce redundancy and create a more streamlined network; finally, KEFBLS employs random Fourier features to map the processed hidden layers, further enhancing the network's feature extraction capabilities, constructing a new learning model. Our experimental results on three UCI regression datasets demonstrate that KEFBLS surpasses other methods in terms of learning efficiency and model performance.

Keywords: Neural Networks and their Applications, Deep Learning, Machine Vision
Abstract: Deep learning techniques have significantly advanced change detection in remote sensing imagery. However, building change detection presents challenges due to the varied appearance of buildings and the complexity of scenes in remote sensing images. Current deep learning-based methods encounter three primary issues. Firstly, CNN-based approaches struggle to model crucial global contextual information essential for remote sensing building images analysis. Transformer-based methods may inadvertently degrade local features. Secondly, traditional attention mechanisms fall short in effectively modeling spatial and spectral features. Thirdly, certain channel attention methods extract excessive redundant information.To address these challenges, this study proposes SFAM-Net, a two-branch hybrid architecture. Our approach initially employs orthogonal methods to minimize redundant information extracted from channels and spaces. Subsequently, we leverage the parallel structure of convolutions and visual transformers to enhance images representation, integrating local features and global representations through cross-attention to better coordinate building and background features. In the CNN and Transformer branches, we adopt spatial-spectral feature coordination and spectral multi-head attention coordination strategies to improve performance in complex scenes. Additionally, we introduce a novel loss function combining edge and center guidance, focusing on changing image edges and centers to enhance sensitivity and accuracy in change area detection. Extensive experiments on widely used LEVIR-CD and WHU-CD datasets validate the effectiveness and efficiency of our network.

Keywords: Active BMIs, BMI Emerging Applications, Other Neurotechnology and Brain-Related Topics
Abstract: By utilizing brain signals, the Brain-Computer Interface (BCI) enables non-muscular communication. In recent decades, BCI systems —particularly speller interfaces— have offered a variety of graphical user interfaces (GUIs). Many attempts had been made to improve the system’s user-friendliness and speller speed. In this paper we present a web-based BCI speller based on Code-Modulated Visual Evoked Potentials (cVEPs), which can be accessed at https://bci-lab.hochschule-rhein-waal.de/en/cvepspeller/. As a result, this web speller is now available to BCI researchers worldwide free of charge, and can be used with variety of own classifier applications. This web speller was successfully tested with the majority of modern web browsers. In the three-step web speller, each character can be selected by navigating through three distinct web interface screens. In this study the web-based cVEP speller was tested and compared to our “local speller” (incorporating the GUI and the signal processing in one application) by seven subjects, who were asked to spell the words “BCI_LAB” and “KLEVE”. All subjects were able to perform the spelling tasks, with a mean accuracy of 95.02% and an average Information Transfer Rate (ITR) of 42.55 bits/min, compared to our “local speller” with a mean accuracy of 93.81% and an average ITR of 48.58 bits/min, respectively. The results showed similar values, confirming the suitability of the suggested web speller for the representation of the cVEP stimuli.

Keywords: BMI Emerging Applications, Active BMIs
Abstract: Motor Imagery-based Brain Computer Interface (BCI) system, that decodes imagined movements from non-invasive electroencephalogram (EEG) are of significant importance, as it can enhance neurorehabilitation and human- computer interaction. Decoding the kinematic parameters of imagined movement is essential to realize BCI systems with higher degrees of freedom of movement and precise control over external effectors. In this work, we propose an efficient algorithm to decode imagined bi-directional movements of hand at two different speeds, slow and fast, from EEG-based BCI. EEG is recorded from fourteen healthy subjects, while they imagined slow and fast movement of their right hand towards the right or left direction. Wavelet-Common Spatial Pattern (WCSP) features and Movement Related Cortical Potential (MRCP) features are extracted from the EEG and a subset of these features are further identified based on the subject-specificity of discriminative subbands and channels. Selected WCSP and MRCP features are then concatenated and used to decode the imagined slow and fast bi-directional movements. Binary classification of the speed-direction pairs resulted in an average classification accuracy of 68.1% across fourteen subjects. To our knowledge, this is the first work addressing decoding of imagined speed and direction of hand movements using EEG-BCI.

Keywords: BMI Emerging Applications, Other Neurotechnology and Brain-Related Topics
Abstract: The burden of cognitive impairment is increasing worldwide and the rehabilitation of patients with such disorders is a major concern. Neurofeedback training (NFT) is emerging as a promising non-pharmacological intervention for enhancing cognition in healthy and cognitive-deficit patients. In this paper, we examine the efficacy of a real-time neurofeedback game using a four-channel consumer-grade Electroencephalography (EEG) system on enhancing overt and covert attention skills in Stroke and Mild cognitive impairment (MCI) patients. The game works based on navigating a car on a computer screen using the player’s attention level computed from EEG signals streamed in real-time. The proposed NFT is conducted across 18 patients (10 Stroke and 8 MCI). The attention score significantly improved from the first session to the final session by 1.25-12.55% in the stroke group and 1.84-44.14% in the MCI group. The experimental results demonstrate that the proposed neurofeedback game using consumer-grade EEG is an effective tool for enhancing overt and covert attention skills in patients with cognitive impairment. To our knowledge, this is the first EEG-Brain Computer Interface study for improving overt and covert attention on the cross-sectional stroke-MCI patients.

Keywords: Other Neurotechnology and Brain-Related Topics, Active BMIs, BMI Emerging Applications
Abstract: Amyotrophic Lateral Sclerosis (ALS) has been a grossly misrepresented end user group when developing coadaptive algorithms for Brain Computer Interfaces (BCI). Researchers have credited this issue to the difficulty of progressing disease in patients with ALS. This non-stationarity reduces accuracy over time. This paper introduces an online model, usable for a BCI using ALS patients data. The automatic coadaptive model effectively decodes 3 class motor imagery (MI) of the left, right hand and rest while adapting to address non-stationarities of Electroencephalography (EEG) over time caused by various factors over the study duration. Adapting Filter bank Common Spatial Pattern (FBCSP) algorithm, where we show it could enable above 70% detection of hand MI in ALS end users longitudinally, previously lacking evidence. The evaluation results demonstrate that the model achieves average accuracies of 72.6% over a 1–2 month period of usage involving 8 ALS patients. This work shows the first auto-adaptive model with ALS patient EEG data providing a stronger incentive for further investigation by setting benchmark models on longitudinal datasets contributing to the solution of multiple challenges in this field.

Keywords: Other Neurotechnology and Brain-Related Topics, BMI Emerging Applications, Active BMIs
Abstract: Post-training quantization (PTQ) is a technique used to optimize and reduce the memory footprint and computational requirements of machine learning models. It has been used primarily for neural networks. For Brain-Computer Interfaces (BCI) that are fully portable and usable in various situations, it is necessary to provide approaches that are lightweight for storage and computation. In this paper, we propose the evaluation of post-training quantization on state-of-the-art approaches in brain-computer interfaces and assess their impact on accuracy. We evaluate the performance of the single-trial detection of event-related potentials representing one major BCI paradigm. The area under the receiver operating characteristic curve drops from 0.861 to 0.825 with PTQ when applied on both spatial filters and the classifier, while reducing the size of the model by about times 15. The results support the conclusion that PTQ can substantially reduce the memory footprint of the models while keeping roughly the same level of accuracy.

Keywords: Passive BMIs, BMI Emerging Applications, Other Neurotechnology and Brain-Related Topics
Abstract: Brain-computer interfaces (BCIs) enable direct interaction between users and computers by decoding brain signals. This study addresses the challenges of detecting P300 event-related potentials in electroencephalograms (EEGs) and integrating these P300 responses for character spelling, particularly within oddball paradigms characterized by uneven P300 distribution, low target probability, and poor signal-to-noise ratio (SNR). This work proposes a weighted ensemble spatio-sequential convolutional neural network (WE-SPSQ-CNN) to improve classification accuracy and SNR by mitigating signal variability for character identification. We evaluated the proposed WE-SPSQ-CNN on dataset II from the BCI Competition III, achieving P300 classification accuracies of 69.7% for subject A and 79.9% for subject B across fifteen epochs. For character recognition, the model achieved average accuracies of 76.5%, 87.5%, and 94.5% with five, ten, and fifteen repetitions, respectively. Our proposed model outperformed state-of-the- art models in the five-repetition and delivered comparable performance in the ten and fifteen repetitions.

Keywords: Intelligent Transportation Systems, Cooperative Systems and Control, Distributed Intelligent Systems
Abstract: Deep reinforcement learning (DRL) based cooperative adaptive cruise control (CACC) of connected and autonomous vehicles (CAVs) shows great potential in improving controllers’ adaptability to real-world dynamic traffic conditions. However, most DRL-based CACCs primarily focus on fixed predecessor following (PF) information flow topology (IFT) while ignoring the heterogeneity of the vehicle. In reality, even though most CACCs separate the platoon by human-driven vehicle (HDV) while considering the HDV as the leading vehicle of the platoon, the heterogeneity of the preceding vehicle with different motion properties (i.e. HDV and CAV) still exists. The adaptability of the DRL-based CACC may be degraded if the above heterogeneity is ignored. To fill this research gap, a DRL-based CACC with preceding vehicle properties is proposed in this study. Specifically, it first designs three typical sub-controllers by considering the applicable IFTs and the properties of preceding vehicles. The proposed DRL-based CACC is established according to the aforementioned designs with the objectives of safety, efficiency, smoothness, and comfort. To enhance the efficiency of transforming acquired data into the learning experience during the exploration process of the DRL algorithm, a periodic update deep deterministic policy gradient (DDPG) algorithm is proposed. The results demonstrate the necessity of considering the preceding vehicle properties in designing the corresponding controller for a specific IFT.

Keywords: System Architecture
Abstract: The index plays an important role in the performance of IoT time series data storage systems. However, the current index designed for HDD or SSD can not adapt to the characteristics of IoT time series data and effectively leverage the performance advantages of NVM. A new hierarchical heterogeneous index is designed for IoT time series storage system based on NVM. The structure is given first and the group consists of several data blocks used to manage IoT time series data. The ordered construction strategy is designed for skip list by the sustained generation of IoT time series data and creating the index for each IoT time series data block group. Meanwhile, a compression and reconstruction strategy for skip list is given to effectively utilize NVM. Then, a TS-Radix tree is presented to index IoT time series data block groups by the temporal characteristics of IoT time series data. Finally, a prototype of it is implemented and YCSB-TS is used to evaluate. The results show that this new index can effectively improve the throughput of random and range queries by up to 262.4%, surpassing the performance of InfluxDB, OpenTSDB, and KairosDB.

Keywords: Electric Vehicles and Electric Vehicle Supply Equipment
Abstract: With the rapid development of Autonomous Rail Rapid Transit technology, lithium-ion battery as its main power source, the research of its charging technology has become particularly important. At present, the charging scheme of lithium-ion battery mainly includes two ways: single high-power charging and multiple low-power charging modules in parallel. However, the single high-power charging scheme has the problems of high cost and low efficiency, and the traditional parallel charging method lacks effective module management strategy, resulting in unbalanced load between modules during charging, affecting charging efficiency and safety. Aiming at the shortcomings of current research, this paper proposes a parallel charging scheme of multiple DC-DC modules based on cooperative control. By designing a closed-loop control system of current inner loop and voltage outer loop, the precise control and cooperative work of parallel modules are realized. The experimental results show that the scheme not only improves the charging efficiency, but also ensures the stability and safety of the charging process, which provides an effective and reliable solution for the lithium-ion battery charging of the intelligent rail train.

Keywords: Intelligent Transportation Systems
Abstract: With rubber wheels instead of steel wheels and no need to be guided by steel rails, Autonomous rail Rapid Transit(ART), is gradually coming into people's lives. However, ART still occupies existing lanes and the relatively small station spacing of ART means that ART needs to be started and stopped frequently, all of which can lead to fluctuations in DC bus voltage during ART operation, making it difficult for loads (such as motors, air conditioners, and sensors) to operate properly. Therefore, this paper proposes the use of predictive set-point modulation to suppress DC bus voltage fluctuations. The predictive set-point modulation method is able to predict the direction of DC bus voltage changes prospectively, and then adjust the voltage preset value to balance the fluctuation of the output voltage, optimizing the closed-loop system's transient dynamic performance. Moreover, since lithium ion battery has high power and high energy consumption, using only one DC-DC circuit can reduce system reliability and cost. Therefore, we propose to use parallel DC-DC modules to balance the excessive power of the battery and verify the feasibility of the proposed method through simulation experiments. The experiments show that the proposed method can effectively suppress the DC bus voltage fluctuation and improve the system reliability.

Keywords: Electric Vehicles and Electric Vehicle Supply Equipment
Abstract: Electric Vehicle (EV) is regarded as the optimal alternative to traditional fuel-powered vehicles. However, the exponential surge in EV charging demand poses challenges in charging infrastructure planning and charging behavior management. The efficacy of traditional Grid-to-Vehicle (G2V) charging mode, which obtains power from the grid, is curtailed by the limited number and uneven distribution of charging facilities, inevitably leading to charging congestion. Instead, the concept of Vehicle-to-Vehicle (V2V) charging has emerged as a spatio-temporally flexible charging mode, which expands EV's role from consumer to provider, forming V2V pairs and utilizing urban Parking Lots (PLs) as charging locations. In this paper, we propose a Hybrid Heterogeneous Modes (HHM)-based EV charging optimization scheme in urban settings. Building upon the G2V mode, we introduce synchronous and asynchronous V2V charging modes as optimization strategies, integrating and exploiting the unique advantages of each mode. We also utilize global EV charging scheduling and comprehensively take four dimensions into consideration (energy trading cost, travel cost, waiting time cost and loss cost), thus achieving flexible mode selection, V2V pairing, and designated charging locations. Simulations confirm the effectiveness of the proposed scheme in reducing total charging costs, optimizing user service experiences, and improving charging facility utilization rates.

Keywords: Autonomous Vehicle, Trust in Autonomous Systems, Fault Monitoring and Diagnosis
Abstract: The development of autonomous and remote-operated driving systems requires extensive stakeholder analyses, requirement engineering, and formalized system descriptions. This is necessary to guarantee the success of the final product after the expensive and time-consuming development phase. To integrate a formalized description of the required abilites of the system, ability graphs have been proposed in the literature. Up to this date, however, this ability graph has only been used to model less complicated driver assistance systems in the literature. This work aims to introduce the value of an ability graph-based description of complex driving systems. This is achieved by successfully demonstrating and discussing a method for constructing a holistic ability graph capable of describing the entirety of abilities required for any driving system.

Keywords: Quality and Reliability Engineering, System Architecture
Abstract: Reed-Solomon coding has been widely adopted to protect data storage against failures in storage systems. Recently, proactive fault tolerance is coming to offer an added protection for data storage coping with the increased device failures. Reliability is critical for storage systems. Due to complex system states and fault tolerance patterns, it is very intricate to analyze the reliability of proactive-tolerance Reed-Solomon storage systems. In this paper, a reliability framework is proposed, which combines event-driven simulation with mathematical model to analyze the reliability of proactive-tolerance Reed-Solomon storage systems. Monte Carlo simulation model simulates storage systems operation according to the failure parameters of different subsystems generated from statistical data. A mathematical model is designed to analyze the probability of data loss under certain concurrent failures. The proposed framework models the impact of various device failures (including permanent device failures, transient failures and correlated failures), network bandwidth, and both accurate predictions and false alarms of proactive fault tolerance on the reliability. The proposed reliability framework can be adapted to different Reed-Solomon codes and system configurations, offering versatility in its application and assist system designers in optimizing trade-offs and comparing schemes, which is beneficial for system design and operation.

Keywords: Consumer and Industrial Applications, Quality and Reliability Engineering, System Modeling and Control
Abstract: Airline pilot training is shifting from traditional quantity-based (hourly-based) training to competency-based training (CBT), and assessing non-technical skills in CBT is complex and essential. Additionally, the captain upgrade training takes more than one year; therefore, a shorter upgrade period with adequate competency is expected. In this study, for captain upgrade training, the information entropy is estimated from the probability distribution of the count data of assessment marker (AM) flags raised per flight, which is raised when each competency component needs improvement. It is found that the information entropy time history calculated from raw data converges into that predicted by the Software Reliability Growth Model (SRGM), which has been found in previous studies as a model to represent the time history of AM flags. Moreover, the local maximum of the entropy time history during a specific period from the start of training correlates with the length of the applicant’s training period. When entropies for each competency are separately calculated, the competencies related to the training period length can be discussed. These results are expected to help estimate the captain upgrade training period and to acquire insight into the training status, including quantitative assessment of non-technical skills.

Keywords: Quality and Reliability Engineering, System Modeling and Control
Abstract: Leveraging the concept of Failure Mode and Effect Analysis (FMEA), we propose a simple and systematic approach, namely Transmission Cause and Effect Analysis (TCEA), to achieve a defined goal of reducing transmission risk through effective preventive and control actions in real-world environments. Specifically, the transmission risk of an infectious disease (e.g., COVID-19) is perceived as a combination of the presence of a transmission agent (e.g., SARS-CoV-2 virus or its variants) and the requisite factors that lead to infection of humans and the associated aftermath of infection. TCEA adopts a causal map to represent all possible transmission risks via a brainstorming process. Next, appropriate preventive and control actions associated with each transmission risk are identified. Similar to FMEA, a Risk Priority Number model with Severity, Occurrence, and Detection ratings is adopted for analysis, prioritization, and decision-making. To demonstrate the usefulness of TCEA, a real-world case study on COVID-19 is conducted. The empirical results indicate that TCEA provide a simple, systematic and easy-to-implement approach to effectively analyze and manage transmission risks of COVID-19 in non-healthcare workplaces.

Keywords: Autonomous Vehicle, Quality and Reliability Engineering, System Modeling and Control
Abstract: The advancement of machine vision systems necessitates efficient and accurate signal reconstruction methods to enhance real-time perception and decision-making capabilities. This paper introduces a Generalized Backtracking Regularization Adaptive Matching Pursuit (GBRAMP) algorithm, designed to reconstruct signals within machine vision systems using compressed sensing techniques. The GBRAMP algorithm improves upon existing methods by incorporating regularization for enhanced atom selection and a backtracking approach to accurately estimate sparsity, addressing the limitations of traditional convex optimization, greedy, and Bayesian reconstruction algorithms. The paper provides a comparative analysis of the GBRAMP algorithm against other prominent reconstruction techniques. Experimental results validate the GBRAMP algorithm's improved performance in terms of both reconstruction accuracy and computational speed, making it a competitive solution for the next generation of machine vision systems.

Keywords: Quality and Reliability Engineering, Manufacturing Automation and Systems
Abstract: A Programmable Logic Controller (PLC) is an essentially domain-specific computer used to control physical equipment and is widely used in industrial control fields. It plays a crucial role in automating complex processes for industrial automation systems, requiring high reliability as code vulnerabilities can potentially lead to disasters. Therefore, vulnerability detection in PLC programs is of significant importance. However, the availability of tools supporting vulnerability detection in PLC programming languages is limited. This paper attempts to improve industrial security from the perspective of code security and proposes a static code analysis approach specifically designed for IEC 61131-3 Structured Text (ST) programs. This approach uses structural pattern matching and symbolic execution technology to identify program defects and improve quality by detecting problematic code structures and potential issues early in the development process, thereby reducing the debugging effort required during developments. Considering the characteristic of periodic loop execution in PLCs, we introduce the loop unwinding technique to collect constraints from subsequent execution cycles for detection purposes. Based on the aforementioned approach, we implement a static code analysis tool, ST-Checker and make a series of evaluations. The experimental results show that this method is feasible and can detect potential defects that existing PLC compilers cannot detect, improving the precision of defect detection with data dependencies.

Keywords: System Modeling and Control, Quality and Reliability Engineering, Manufacturing Automation and Systems
Abstract: In this paper, we define reduced dimensional predictors in latent dynamic systems as contrast to the traditional full-dimensional predictor models. Then the estimation of the new latent vector autoregressive model is developed with an objective to maximize the predicted variance for a given number of latent variables. A new dynamic predictive monitoring index that accounts for variations in the prediction residual and the predictor is developed. The residuals are modeled with a subsequent principal component analysis and a comprehensive monitoring method is developed to detect abnormal situations in industrial and operational systems. The new algorithm is tested on a simple closed-loop control system and the revamped Tennessee Eastman simulated process to show its effectiveness compared to other state of the art methods.

Keywords: Deep Learning, Neural Networks and their Applications
Abstract: Multimodal sentiment analysis integrates various modalities of information to collectively inform decision-making processes. Previous studies often treat different modal features equally or emphasize textual information as the primary consideration. However, when the modalities in the sample contain different sentiment information, these methods may not be able to effectively deal with this situation. To solve this problem, we propose a multimodal sentiment analysis model focusing on each modality (MFM). In this paper, we separately integrate each modality as a primary modality interacting with other secondary modal information so that each modality can play a leading role. In addition, we use shared mask in modal interaction to capture important information in the secondary modality related to the primary modality, and improve the effectiveness of the information interaction process. The model is evaluated against baseline models using the MOSI and MOSEI multimodal sentiment analysis datasets. The experimental results show that the model achieves better performance, thereby validating its effectiveness in multimodal sentiment analysis tasks.

Keywords: Transfer Learning, Representation Learning, Information Assurance and Intelligence
Abstract: Federated Domain Generalization (FedDG) aims to learn a generalizable model from source domains while preserving data privacy. However, several challenges exist: (1) effective knowledge transfer across diverse modalities like text or audio has not been achieved; (2) the presence of irrelevant or malicious source domains can cause negative transfer due to the lack of data quality governance; (3) existing methods use Shapley Value for contribution calculation, which has high computational complexity and is only suitable for scenarios with a small number of clients. To address these issues, we propose FedVFGL, a novel FedDG framework. FedVFGL introduces a global objective with personalized feature calibration. Its key innovation is a dynamic aggregation scheme that adjusts each domain's contribution based on relevance and reliability. Theoretical analysis demonstrates that this reweighting can improve generalization bounds over traditional FedDG. FedVFGL is a versatile, robust algorithm that integrates with various federated techniques and mitigates data poisoning. Extensive experiments demonstrate FedVFGL's superior performance over state-of-the-art FedDG methods, and its complementary benefits when combined with existing approaches. The FedVFGL source code is publicly available at: https://github.com/scnu-kevinkong/Fedvfgl.

Keywords: Multimedia Computation, Neural Networks and their Applications, Deep Learning
Abstract: Multimodal Named Entity Recognition (MNER) is a task that leverages multimodal information (such as text and images) to identify named entities within social media text. Traditional MNER methods primarily rely on simple interactions between text annotations and visual features, thus overlooking the specific correspondence between text and visual objects. Additionally, irrelevant visual noise may interfere with the final recognition results. Therefore,this paper proposes a Entity label-guided Graph Fusion Multi-modal Named Entity Recog nition approach(ELGF), which utilizes pre-defined entity label information from input text as a bridge. Firstly, entity label detection tasks are employed to obtain entity label information. Then, the entity label information is utilized as a bridge between the two modalities to construct a multimodal interaction graph. This graph is inputted into a graph neural network, where attention and gate mechanisms are applied to interactively fuse multimodal information. Finally, a Conditional Random Field (CRF) decoding is used to predict the final MNER label sequence.Extensive experimental results demonstrate that compared to mainstream methods, the proposed model achieves competitive recognition accuracy on public datasets.

Keywords: Representation Learning, Deep Learning, Machine Learning
Abstract: Graph Neural Networks (GNNs) often assume strong homophily for graph classification, seldom considering heterophily, which means connected nodes tend to have different class labels and dissimilar features. In real-world scenarios, graphs may have nodes that exhibit both homophily and heterophily. Failing to generalize to this setting makes many GNNs underperform in graph classification. In this paper, we address this limitation by identifying three effective designs and develop a novel GNN architecture called IHGNN (Incorporating Heterophily into Graph Neural Networks). These designs include the combination of integration and separation of the ego- and neighbor-embeddings of nodes, adaptive aggregation of node embeddings from different layers, and differentiation between different node embeddings for constructing the graph-level readout function. We empirically validate IHGNN on various graph datasets and demonstrate that it outperforms the state-of-the-art GNNs for graph classification.

Keywords: Biometric Systems and Bioinformatics, Deep Learning, AI and Applications
Abstract: With the widespread use of smartphones, more and more private information is stored on the phone, and the loss or theft of the phone can lead to data leakage, theft of property, and other problems. Traditional active authentication methods based on passwords, faces, fingerprints, etc. authenticate only once at login, which still has some hidden dangers. Authentication methods based on behavioral biometrics such as walking gait, touch screen, keystroke, etc. can provide implicit and continuous authentication services, and thus have been widely studied recently. Considering the limited computational resources of smart devices, we first introduce the state-space model with linear model complexity, i.e., Mamba, to extract deep feature representations from raw signal sequences in the sensor-based authentication task. Then a series of auxiliary information from different domains are proposed and fused to the deep features to enhance the consistent semantic information and ultimately improve the discriminative ability of the model. Our model A-Mamba, i.e., Auxiliary-Mamba, is experimented on two public datasets. The experimental results show that our model outperforms existing approaches.

Keywords: Deep Learning, Machine Learning, Machine Vision
Abstract: With the rapid advancement of autonomous driving technology, the demand for faster and more accurate object detection frameworks has become increasingly urgent. Recently, many deep learning-based object detectors have demonstrated impressive performance in real-time driving applications. However, the detection of small objects such as traffic signs remains challenging due to the complex nature of these objects. This paper proposes a transformer-based detector TS-DETR to improve the accuracy of small object detection in autonomous driving systems. We introduce a constrained decoder structure to focus the model's attention on the predicted boxes. Additionally, a content-aware deformable cross-attention mechanism is proposed to obtain more comprehensive attention weights. Experimental results on challenging public datasets such as TT100K and CCTSDB2021 demonstrate that our approach achieves significant performance improvements with only a minimal increase in the number of parameters compared to existing algorithms.

Keywords: Information Systems for Design and Marketing, Human-Machine Cooperation and Systems, Human Performance Modeling
Abstract: Peer review is an integral part of academic publi- cation necessary to maintain high standards and novelty of pub- lished research. Despite its importance, peer reviewers are rarely provided incentives, leading to journals having difficulty finding reviewers inclined to accept invitations and submit reviews on time. This paper proposes a Blockchain-based Anonymous Reviewer Incentive Token (BARIT) to incentivize peer reviewers. BARIT introduces flexible incentive schemes to provide both recognition and tangible benefits for the reviewers’ contribution while preserving the anonymity of reviewers. Using blockchain technology to record reward tokens ensures their permanence and acceptance across different publishers. Incentive models are designed to encourage the involvement of researchers as reviewers, reduce reviewer refusal rates, and prompt the timely submission of review reports.

Keywords: Human-Machine Interaction, Human-Machine Cooperation and Systems, Human-Computer Interaction
Abstract: Recent Large Language Models (LLMs) demonstrate problem-solving capabilities suitable for educational use. This paper investigates using LLMs to create virtual agents that mimic student behavior and interact with learning platforms. Testing a modest-sized LLM on an Intelligent Tutoring System for word problem-solving revealed that the LLM could fully solve 92% of single-step problems, although their performance decreased to 14% when attempting more complex problems.

Keywords: Machine Learning, Machine Vision, Deep Learning
Abstract: The reconstruction of dense point clouds is an important foundation for downstream applications, such as object detection, semantic classification and surface reconstruction. Current methods focus on dense point cloud reconstruction for objects but neglect the whole scene. To address this issue, the reconstruction of dense point clouds supervised by inverse stereo matching (IS-Dense) is proposed. In detail, the Transformer model is first used to extract deep features form the point clouds. Second, point cloud features are expanded through the base upsampler. Ultimately, the point clouds would be coordinated following the feature expansion. Due to the uneven distribution of point clouds in the whole scene, some gaps and anomalies are presented in the data. Therefore, a point location refinement module supervised by inverse stereo matching is designed to solve this problem. For this module, the key is to utilize the reconstructed dense point clouds and the right image to estimate the left image. Supervised by real left images, the reconstructed dense point clouds are precise and even-distributed. The experimental results prove the superiority of the proposed method over current methods, especially for the whole scene.

Keywords: Deep Learning, Neural Networks and their Applications, AI and Applications
Abstract: Tables play a crucial role in both documents and daily life, thus sparking significant interest in the research of automatic table structure recognition(TSR). Recent methods primarily achieve recognition by predicting makeup sequences or adjacency relationships between table cells. However, these methods have some limitations in applications. The former requires additional computation of heuristic rules to recover the structure of the table, which not only increases the computational complexity but also may affect the recognition accuracy. The later relies on huge training data and inefficiency decoders, which not only raises the cost but also limits their application in real-time or large-scale data processing scenarios. At the same time, they often ignore the significance of cell logical location and fail to effectively utilize the rich text information that naturally exists in the table. In this paper, we propose a new framework called VTLL to solve the problem. The method extracts visual and textual features to do adaptive feature fusion, and we also introduce a cascading regressors to predict the fused features multiple times, while combining intra-cell and inter-cell losses. As for the regression header part, we use a mixed context aggregator to understand the inter-cell relationships. We evaluated the proposed methods on several public datasets, experiment results demonstrate that VTLL performs better.

Keywords: Information Assurance and Intelligence, Computational Intelligence in Information, Cybernetics for Informatics
Abstract: 在数字社会时代，生成频率 多模态数据正在迅速增加。区块链， 被公认为值得信赖的分布式数据库技术， 为可信存储和 高效管理多模态数据。然而 区块链系统仅支持使用事务的查询 哈希值作为关键字，不能直接利用 多模态数据的内容特征，导致一般 查询效率低。为了解决这个问题，本文 介绍了一种高效的多层索引方法 用于多模态数据查询的区块链。它建立了一个 链上和链下数据之间的有效映射 通过可验证的链上和链下协同 存储体系结构。该论文还提出了多层 位图块索引 （MBBI） 和杜鹃默克尔树 （C-MT） 到 优化查询流程。实验结果 证明这种方法不仅确保了 链上元数据的一致性和完整性，以及 链下还显著提升了效率 多模态数据查询。这

Keywords: Machine Learning, Neural Networks and their Applications, Deep Learning
Abstract: Multi-label learning and continual multi-label learning are crucial challenges in machine learning, particularly in handling complex data with multiple overlapping labels over time. Recent research works try to tackle the effect of data imbalance as it makes multi-label learning more challening. This work introduces an adaptive margin spiking neural network (AM-SNN) architecture coupled with a novel imbalance-sensitive loss function designed to enhance robustness against class imbalance in these settings. AM-SNN employs two output layers: one for predictions and another for margin values, with a unique loss function leveraging cosine similarity between predictions and ground truths to improve confidence in the model's predictions. Experiments show that AM-SNNs trained with the proposed loss function outperform state-of-the-art loss functions on metrics such as the imbalance-weighted F1 score and the F1 score for the most imbalanced class on several multi-label learning datasets. In continual multi-label learning, AM-SNNs surpass Bipolar SNNs and the CIFDM benchmark on large datasets - Birds, Human, and Eukaryote.

Keywords: Quality and Reliability Engineering
Abstract: Smart contracts usually hold a large amount of digital assets, which can cause substantial losses if these contracts have vulnerabilities. Thus, it is essential to adequately detect possible vulnerabilities in smart contracts before deployment. There are many types of vulnerabilities in smart contracts, and different detection methods have their own unique advantages, some vulnerabilities may be more suitable for expert rule-based methods, while some vulnerabilities are more suitable for deep learning-based methods. A single detection method usually fails to fully use its ability to detect vulnerabilities. To address the above problems, we propose a composite approach named CDE-VD (Combining Deep Leaning and Expert Rule for Smart Contract Vulnerability Detection) to improve the performance of vulnerability detection. The method divides smart contract samples into deep learning-prone samples and expert rule-prone samples by classifying them before detection, and extracts expert rule features to train the smart contract detection method classifier to predict the category of the samples under analysis, then selects the suitable method for detection. The experimental results show that the vulnerability detection performance of CDE-VD outperforms that of single detection methods. Compared with the SOTA method MANDO, CDE-VD achieves average improvements of 3.22%, 2.32%, 9.25%, and 6.54% in terms of the Accuracy, Precision, Recall, and F_1-score for five categories of vulnerabilities such as access control and time manipulation, respectively, which indicates that category prediction of the smart contract samples could improve vulnerability detection performance.

Keywords: Application of Artificial Intelligence, Deep Learning, Information Assurance and Intelligence
Abstract: Recent advancements in large language models (LLMs) highlight their fluency in generating responses to diverse prompts. However, these models sometimes generate plausible yet incorrect "hallucinated" facts, undermining trust. A frequent but often overlooked cause of such errors is the use of poorly structured or vague prompts by users, leading LLMs to base responses on assumed rather than actual intentions. To mitigate hallucinations induced by these ill-formed prompts, we introduce Curative Prompt Refinement (CPR), a plug-and-play framework for curative prompt refinement that 1) cleans ill-formed prompts, and 2) generates additional informative task descriptions to align the intention of the user and the prompt using a fine-tuned small language model. When applied to language models, we discover that CPR significantly increases the quality of generation while also mitigating hallucination. Empirical studies show that prompts with CPR applied achieves over a 90% win rate over the original prompts without any external knowledge.

Keywords: Application of Artificial Intelligence, Deep Learning, Machine Vision
Abstract: Text-to-image(T2I) synthesis aims to generate semantically consistent images with texts. Currently, existing methods merely use the shallow semantics of the text to crudely guide image generation. They cannot fully integrate rich textual semantics with image features, leading to an inability to control image generation through text finely. To address this issue, We propose a GAN-based method named MVESF (Multi-View Enhanced Semantic Fusion), which enhances the semantic fusion of text and images from multiple perspectives for fine-grained controllable text-to-image synthesis. In multi-view, we introduce Multi-domain Semantic Guidance, Local Semantic Attention, and Visual-textual Consistency Loss to enhance the semantic fusion of text and images in image generation, image discrimination, and image supervision, respectively. Our method promotes the consistent alignment between text and images, allowing for fine-grained variations in the generated images when subtle changes in the input text without affecting unrelated regions. Extensive experiments have demonstrated the effectiveness of our approach.

Keywords: Application of Artificial Intelligence, Deep Learning, AI and Applications
Abstract: Electricity theft severely impairs the economic benefits to businesses and endangers public safety. In recent years, deep learning models for automated electricity theft detection have been advancing rapidly. However, the pattern of electricity usage data exhibits complex dependencies in cases of electricity theft, posing significant challenges for accurate detection. To cope with this difficulty, we propose a Channel Independent Attention Network (CIAN) for electricity theft detection. In particular, electricity consumption data is firstly processed by a sequence patchfication strategy, which enables dependency capturing over suitable time intervals. A parallel structure is then constructed, where a self-attention based channel independent encoder and a mix convolution module are cooperated for learning global and local feature patterns, respectively. Lastly, the features are fused to predict whether the electricity consumption sequence data involves electricity theft. Experimental results on a real-world electricity dataset demonstrate that the proposed method surpasses other competitive electricity theft detection approaches across various evaluation metrics.

Keywords: Application of Artificial Intelligence, AI and Applications, Deep Learning
Abstract: Tibetan-Chinese machine translation has become a focal point of interest within the Tibetan community due to its importance for effective communication and cultural preservation. Although technological advancements have been made, current Tibetan-Chinese translation systems still fall short of satisfactory performance. Moreover, evaluating these systems using a publicly available and reliable benchmark dataset remains problematic. To address these challenges, we propose an enhanced Tibetan-Chinese machine translation framework. This framework includes data filtering techniques and utilizes the semantic knowledge of the state-of-the-art Chinese cross-lingual model, CINO. After preprocessing, we trained our model using the Transformer architecture and found that our approach significantly improves translation performance. Additionally, we have developed a high-quality, expert-reviewed test dataset to support community evaluations of translation systems. The details of our experimental model and the test dataset can be accessed at footnote{https://github.com/UTibetNLP/Ti-ChTrans}.

Keywords: Application of Artificial Intelligence, Deep Learning, Image Processing and Pattern Recognition
Abstract: Aiming at the problem that the traditional detection method has low efficiency and accuracy due to the fact that there are many key component targets to be inspected by unmanned aerial vehicle (UAV) during power inspection, and the shape difference is large, and the image quality is not high, a transmission line abnormal target detection method based on improved YOLOv8 and super-resolution reconstruction is proposed. Firstly, the super-resolution reconstruction algorithm is used to reconstruct the abnormal image to improve the clarity and enrich the characteristic information contained in the image. On this basis, the improved YOLOv8 network is used to detect the defects in the inspection image. The CBAM attention mechanism is fused in the Bottleneck part of the C2f module to strengthen the model's ability to locate the target; In order to further improve the detection ability of small targets in patrol inspection, a small target detection layer is added to make the network pay more attention to the detection of small targets.Finally, in order to be deployed to the edge devices, the original convolution in the network is modified to a lightweight convolution GhostConv to reduce the number of model parameters. The experimental results show that the proposed method can accurately detect abnormal defects of transmission line components on the basis of improving the quality of inspection images. mAP is improved by 3.7% and the number of model parameters and calculation amount are greatly reduced, which reflects the effectiveness of the algorithm, and it has stronger extraction ability and robustness for subtle defect targets, meeting the detection requirements of power inspection.

Keywords: AI and Applications, Application of Artificial Intelligence, Media Computing
Abstract: Generative linguistic steganography aims to embed information into natural language texts to achieve covert transmission. However, currently in most approaches based on subword-supporting language models, the extraction process relies on tokenizing steganographic texts into tokens, which could cause segmentation ambiguity, leading to false results or failures of extraction finally. Despite several existing countermeasures (or disambiguation) that have been proposed, they are based on removing tokens of candidate pools, which render them incompatible from the sights of keeping imperceptibility, potentially incurring safety risks. To avoid it, we focus on tackling segmentation ambiguity with near-integrity of candidate pools. In this paper, we propose a near-imperceptible disambiguating approach via verification for generative linguistic steganography. First, this paper draws an all-case extraction method to obtain possible true extracted results. Further, length verification and checksum verification are presented to filter wrong extracted results caused by segmentation ambiguity. Experiments show that our disambiguating approach outperforms the existing disambiguating approaches, on various criteria, including about 23.49% higher embedding capacity, about 23.46% higher imperceptibility and about 5.73% anti-steganalysis capacity of steganographic texts.

Keywords: Neural Networks and their Applications, Machine Vision, Machine Learning
Abstract: Abstract—Traditional autofocus methods search for the optimal focal distance (FD) by evaluating image quality from focal stacks, resulting in time-consuming focusing processes. Recently, deep learning has being adopted for single-shot autofocus methods, which can predict the optimal FD directly from a single input image. However, these methods often suffer from low prediction accuracy due to the lack of global features and structured global supervisory information, as they rely solely on the image’s region of interest (ROI) as input and a single value for supervision. We propose a deep learning network named MPFS, which takes a full-frame photograph as input and uses the optimal focal distance per pixel for supervision, this method effectively addresses the issues of missing global features and insufficient supervisory information by leveraging these enhancements. Additionally, the network integrates current camera focal distance information to mitigate the scale ambiguity caused by the lack of absolute scale information. To validate the effectiveness of the proposed method, we designed an experiment using a dataset annotated with optimal FD per pixel. Experimental results on this dataset indicate that our approach achieves a 0.22 decrease in the Mean Absolute Error (Mae) metric compared to the state-of-the-art models, with improvements of 0.02 and 0.004 in d1 and d2 metrics.

Keywords: Neural Networks and their Applications, Machine Learning, Image Processing and Pattern Recognition
Abstract: Plant diseases are one of the factors that compromise food production goals, and the characteristics and climate of each production region influence them. Tomatoes are one of the world's most consumed vegetables and are widely affected by various diseases. However, tomato cultivation in greenhouses allows its continuous production. In this context, this research work focuses on the problem of identifying diseases in tomato cultivation scenarios in greenhouses. For this study, we created new datasets with two image sizes: the Tomato Leaf Image Dataset (TLID) with image sizes of 256x256 pixels and 15,256 images, and the Patch Tomato Leaf Image Dataset (PTLID) with patch sizes of 32x32 pixels and 227,218 images. Both datasets comprise seven classes, including four types of diseases, two combinations of diseases on the same leaf, and the healthy leaf. Machine Learning techniques have been widely used to identify plant diseases. This work presents two machine learning methods tested with both datasets. In the proposed models, we combine three convolutional neural networks, a customized CNN, VGG19, and Resnet50, and two voting classification methods using Hard and Soft decisions. The evaluation performed on the datasets showed that when the patches are used, the results improve significantly, reaching an accuracy of 90.48%. This technique makes it possible to identify the stage of the disease.

Keywords: Deep Learning, Machine Learning, Big Data Computing,
Abstract: In location-based services (LBS), providing accurate trip recommendations is a challenge due to the diverse trip preferences of users and the complexity of their transfer behaviors. Previous trip recommendation studies neglect the effect of the following POIs on the previously visited ones, called a reverse effect. To address this problem, this study builds a Graph-convolutional-network-based Double-layer Bidirectional Trip Recommendation (GDB-TR) model. It utilizes a heterogeneous graph to model users’ check-in trajectories, incorporating spatial and temporal information. Subsequently, subgraphs are extracted, and adjacency matrices are constructed to represent the relationships within each subgraph. By fusing these matrices through a neural network, GDB-TR obtains vector representations for both point-of-interests (POIs) and POI categories. The core of GDB-TR is a double-layer bidirectional neural network. This network comprises two layers: one for describing POIs and the other for POI categories. Bidirectional computation is conducted between the initial and destination nodes, with a forward computation capturing the influence of preceding POIs or categories on following ones, and reverse computation evaluating this influence in the opposite direction. Finally, we conduct experiments on five popular real-world datasets and the results show the superiority of GDB-TR over existing baseline models, measured by F1 and pairs-F1 metrics.

Keywords: Deep Learning, Neural Networks and their Applications, Image Processing and Pattern Recognition
Abstract: Self-supervised learning has demonstrated impressive performance in speech tasks, yet there remains ample opportunity for advancement in the realm of speech enhancement research. In addressing speech tasks, confining the attention mechanism solely to the temporal dimension poses limitations in effectively focusing on critical speech features. Taking into account the aforementioned issues, our study introduces a novel speech enhancement framework, HFSDA, which skillfully integrates heterogeneous spatial features and incorporates a dual-dimension attention mechanism to significantly enhance speech clarity and quality in noisy environments. By leveraging self-supervised learning embeddings in tandem with Short-Time Fourier Transform (STFT) spectrogram features, our model excels at capturing both high-level semantic information and detailed spectral data, enabling a more thorough analysis and refinement of speech signals. Furthermore, we employ the innovative Omni-dimensional Dynamic Convolution (ODConv) technology within the spectrogram input branch, enabling enhanced extraction and integration of crucial information across multiple dimensions. Additionally, we refine the Conformer model by enhancing its feature extraction capabilities not only in the temporal dimension but also across the spectral domain. Extensive experiments on the VCTK-DEMAND dataset show that HFSDA is comparable to existing state-of-the-art models, confirming the validity of our approach.

Keywords: Transfer Learning, Neural Networks and their Applications, Deep Learning
Abstract: 迁移学习已成为深层次的关键技术 学习，在工业界和学术界广泛采用 开发定制模型，特别是针对特定和 下游任务解决。尽管 迁移学习，目标模型可以轻松继承 在学习过程中源模型的缺陷， 例如容易受到后门攻击和对抗性攻击 攻击。因此，这项工作提出了一种新的方法， 迭代剪枝学习方法 （IPLA），可减少 转移过程中潜在缺陷的继承 学习过程。为了减少对 攻击并提高目标模型IPLA的鲁棒性 评估源模型中权重的重要性 并保留对目标任务至关重要的任务，然后 通过迭代修剪来修剪多余的权重 过程。实验在 4 个数据集上进行，超过 2 个数据集 骨干源模型。结果表明 所提方法性能令人满意。

Keywords: Neural Networks and their Applications, Deep Learning, Optimization and Self-Organization Approaches
Abstract: Pickup and Delivery Problem with Time Windows (PDPTW) is a prevalent research direction in modern logistics transportation. In this challenging problem, customers are divided into pickup nodes and delivery nodes, and vehicles must first serve each pickup node before proceeding to its corresponding delivery node. Moreover, the hard time window constraint presents an obstacle for the existing learning-to-construct methods. Hence, this paper proposes a novel learning-to-construct approach based on node-pair insertion to address the complex time window constraint. It involves predicting the insertion point for the next node pair within the current partial solution and ensuring constraint adherence. We enhance the context information for the decoder to produce better solutions. The experimental results verify that the proposed approach can construct high-quality solutions in a very short period of time.

Keywords: Deep Learning, Image Processing and Pattern Recognition, Transfer Learning
Abstract: Medical images from different healthcare centers exhibit varied data distributions, posing significant challenges for adapting lung nodule detection due to the domain shift between training and application phases. Traditional unsupervised domain adaptive detection methods often struggle with this shift, leading to suboptimal outcomes. To overcome these challenges, we introduce a novel domain adaptive approach for lung nodule detection that leverages mean teacher self-training and contrastive learning. First, we propose a hierarchical contrastive learning strategy to refine nodule representations and enhance the distinction between nodules and background. Second, we introduce a nodule-level domain-invariant feature learning (NDL) module to capture domain-invariant features through adversarial learning across different domains. Additionally, we propose a new annotated dataset of X-ray images to aid in advancing lung nodule detection research. Extensive experiments conducted on multiple X-ray datasets demonstrate the efficacy of our approach in mitigating domain shift impacts.

Keywords: Image Processing and Pattern Recognition, Information Assurance and Intelligence, Neural Networks and their Applications
Abstract: Deepfake技术的普及提高了 识别真假面孔的挑战。而 检测方法已经存在，其中大多数是被动的 取证并面临普遍性的挑战，以及 迁移。目前，一些研究试图保护 通过预先插入不可见信息来获得原始图像。 但是，在形象方面仍然存在不足 由于信息而产生的质量和信息稳健性 嵌入，即水印。因此，我们采用 感知哈希编码的鲁棒性，并将其与 提出主动的信息隐藏技术 Deepfake检测解决方案，在此简称PWPH 纸。我们的方法简单而高效：首先， 包含人脸的图像分为两部分：FA（人脸 area）和NFA（非面部区域）。感知哈希代码是 从非面部区域 （NFA） 生成。然后，哈希 代码作为水印嵌入到 FA 中。在 提取阶段，我们使用与编码器相同的方法来 从 FA 中检索嵌入的水印。水印਷

Keywords: Image Processing and Pattern Recognition, Machine Vision, Deep Learning
Abstract: With the development of generative models, unpaired image translation has made remarkable progress. It aims to translate an image from one domain to another domain while remaining the content information of the input image. However, existing methods still suffer from semantic inconsistency during the translation process. In this paper, we propose a novel unpaired image translation framework to improve semantic consistency via high and low frequency guidance. Our key idea is to divide images into high and low frequency components. Based on frequency representation of images, we design a new high-frequency encoding module that remains content semantic information from the original domain. We also present a novel style consistency regularization on low-frequency components, which enhances style consistency of output images. Our approach achieves sota performance on multiple unpaired image translation tasks. Extensive experiments prove our method significantly improves semantic consistency that remains content semantics from the original images and generates style-harmonious images.

Keywords: Application of Artificial Intelligence, Image Processing and Pattern Recognition, Deep Learning
Abstract: When it comes to classifying child sexual abuse images, managing similar inter-class correlations and diverse intra-class correlations poses a significant challenge. Vision transformer models, unlike conventional deep convolutional network models, leverage a self-attention mechanism to capture global interactions among contextual local elements. This allows them to navigate through image patches effectively, avoiding incorrect correlations and reducing ambiguity in attention maps, thus proving their efficacy in computer vision tasks. Rather than directly analyzing child sexual abuse data, we constructed two datasets: one comprising clean and pornographic images and another with three classes, which additionally include images indicative of pornography, sourced from Reddit and Google Open Images data. In our experiments, we also employ an adult content image benchmark dataset. These datasets served as a basis for assessing the performance of vision transformer models in pornographic image classification. In our study, we conducted a comparative analysis between various popular vision transformer models and traditional pre-trained ResNet models. Furthermore, we compared them with established methods for sensitive image detection such as attention and metric learning based CNN and Bumble. The findings demonstrated that vision transformer networks surpassed the benchmark pre-trained models, showcasing their superior classification and detection capabilities in this task.

Keywords: Image Processing and Pattern Recognition, Deep Learning, Neural Networks and their Applications
Abstract: Text detection is a task of great significance in different scenarios, which has wide applications for downstream tasks such as text recognition and text retrieval. Varieties of text detection methods have been proposed to solve this problem in natural scenes and have achieved good results. However, these methods cannot get satisfied performance in industrial scenes for various interferences caused by the industrial environment like the image noise, background material texture and low contrast. To deal with these challenges, we first propose a contour modeling algorithm based on graph Fourier transform mapping to represent arbitrary shaped text contours. A refined fast Fourier convolutional network module with ability of spectral-domain sensing is also intruduced to enhance text feature and suppress interference. Based on these two components, we construct a novel network to achieve accurate industrial scene text detection. Quantitative evaluations are conducted on benchmark datasets MPSC and IcText, experimental results show that our method obtains the state-of-the-art detection accuracy.

Keywords: Deep Learning, Image Processing and Pattern Recognition, Application of Artificial Intelligence
Abstract: This paper presents two novel approaches for improving skeleton-based action recognition using Graph Convolutional Networks (GCN) and Convolutional Neural Networks (CNN). In the first approach, we combine GCN and CNN streams that process position and velocity features to improve classification accuracy. In the second approach, we use GCN as an embedding layer for support network CNN to extract features from skeleton data, which significantly improves recognition accuracy. Our experiments on the JHMDB dataset demonstrate that our approaches outperform state-of-the-art methods while using significantly fewer parameters. Additionally, we extended our evaluation to the Kinetics-400 dataset, where our methods showed comparable results with considerably lower model complexity. Our work contributes to the development of more efficient and robust action recognition models.

Keywords: Brain-Computer Interfaces, Human-Machine Interaction, Human-Machine Interface
Abstract: To address the effect of both intra- and inter-subject variability in the EEG-based motor-imagery classification, adaptive schemes have been proposed whereby the pre-trained model. However, collection and labelling of additional target data is resource intensive. Furthermore, there may exist significant signal variations across time especially across multiple recording sessions which can result in model performance deterioration following adaptation. This introduces another challenging problem to training classifiers as they are typically unable to automatically determine which data is suitable for fine-tuning purposes, leading to some subjects facing performance drops even after adaptation. To address the data scarcity and variability in adaptive performance, we propose a novel machine-relearning Siamese architecture which utilizes few samples of unlabeled evaluation data to perform data efficient model re-learning. Machine re-learning optimally selects a sub-section of previously known data to update the model parameters. This is implemented via the use of comparative contrastive loss between the Gaussian distributions of target and known data to perform data selection. The highest subject-independent performance achieved an average (N=54) accuracy of 86.63% (±11.79%) without additional data and 88.23% (±10.36%) when utilizing a single unlabeled supplementary data for two-class motor imagery. The previous best accuracy on this dataset is 85.90% (±11.20%) using the best-known method in the literature. Therefore, significantly superior adaptation performance can be achieved while utilizing lesser amount of information from the target subject through reducing EEG feature variability in the training and fine-tuning sets. Codes may be found at: https://github.com/NgHanWei/EEG_Relearning

Keywords: Brain-Computer Interfaces, Human-Machine Interaction, Medical Informatics
Abstract: Research on transcranial magnetic stimulation (TMS) combined with encephalography feedback (EEG-TMS) has shown that the phase of the sensorimotor mu rhythm is predictive of corticospinal excitability. Thus, if the subject-specific optimal phase is known, stimulation can be timed to be more efficient. In this paper, we present a closed-loop algorithm to determine the phase linked to the highest excitability with as few trials as possible. We used Bayesian optimization with different configurations as an automated, online search tool in an EEG-TMS simulation experiment. From a sample of 38 healthy participants (25 f, 18 m), we selected all participants with a significant single-subject phase effect (n = 5) for the simulation. We then simulated 1000 experimental sessions per participant where we used Bayesian optimization to find the optimal phase. We tested two objective functions: Fitting a sinusoid in Bayesian linear regression, or Gaussian Process regression. We additionally tested adaptive sampling using a knowledge gradient as the acquisition function compared with random sampling. We evaluated the algorithm’s performance in a fast optimization (100 trials) and a long-term optimization (1000 trials). We found that for fast optimization, the Bayesian linear regression in combination with adaptive sampling gives the best results with a mean phase location accuracy of 79 % after 100 trials. With either sampling approach, Bayesian linear regression performs better than Gaussian Process regression in the fast optimization. In the long-term optimization, Bayesian regression with random sampling shows the best trajectory, with a rather steep improvement and good final performance of 87 % mean phase location accuracy. In summary, we could show the suitability of closed-loop Bayesian optimization for phase selection. We show that we can increase the speed and accuracy by using prior knowledge about the expected function shape compared with traditional Bayesian optimization with Gaussian Process regression.

Keywords: Human-Machine Interface, Wearable Computing, Assistive Technology
Abstract: Cardiac dysrhythmias pose a significant risk of sudden cardiac death, necessitating prompt detection and treatment. While wearable ECG devices offer real-time monitoring capabilities, current methods for identifying dysrhythmias are often time-consuming. In this paper, we present a novel low-complexity RR interval (RRI)-based Cardiac Dysrhythmia Indicator (CDI) aimed at efficiently detecting dysrhythmias. The CDI architecture comprises two principal components: the Slow-Fast-Normal Indicator Circuit (SFNIC) and the Normal-Abnormal Indicator Circuit (NAIC). SFNIC monitors dysrhythmias at each RRI, while NAIC evaluates dysrhythmia status at regular intervals, indicating normal or abnormal states every 64 RRIs (64 bpm is near to normal pulse rate). Through extensive simulations conducted using 0.18 μm CMOS processes, with mixed-signal circuits operating within the (0.9-1.8) volt range, our design demonstrates successful recognition of various ECG rhythms. These findings suggest that the proposed CDI design holds promise for integration into wearable ECG devices, enabling real-time dysrhythmia detection and timely intervention.

Keywords: Human-Machine Interaction, Design Methods, Assistive Technology
Abstract: Automatic self-checkout based on computer vision is gaining popularity in the field of retail industry, due to the convenience for customers and manpower saving. Thus, retail product detection is vital important in the process of automatic checkout. The task of product detection based on single camera is still challenging, like (1) holding a variety of different products in one or both hands, (2) products appearance, (3) intentional fraudulent checkout practices. In this paper, we introduce a third branch on ordinary detectors to predict the occlusion layer of a product and then adopt occlusion layer aware non-max suppression (OLA-NMS) to depress false positives while keeping detection rate. Furthermore, IoU-activate loss is adopted by considering location information in the classification loss. Our third contribution is that we have collected a large-scale of retail checkout images for the target of self-checkout monitoring (SCOM), since there is no dataset or benchmark available for retail product detection under occlusion. Experiments are conducted on SCOM dataset to demonstrate the effectiveness of the proposed method.

Keywords: Human-Machine Cooperation and Systems, Resilience Engineering, Human Factors
Abstract: Faults in the technical process of automated production systems (aPS) are challenging to detect, report, and recover accurately. Missing physical equipment, lacking detection mechanisms in the control software, and insufficient training and experience of human operators lead to an often delayed and imprecise detection of a fault’s root cause. Since fault detection and recovery typically rely heavily on human involvement, there is also a risk of operators performing only partial repairs or inadvertently introducing new faults. Thereby, even slight deviations in human-machine interac-tions may affect process outcomes unpredictably, which so far cannot be modelled in field-level automation. The Functional Resonance Analysis Method (FRAM) is well-suited to address functional relations in such complex socio-technical systems and accurately represent resource requirements, preconditions, control, and time constraints, making it a promising tool to methodically investigate aPS faults. Yet, the FRAM has barely been applied in this domain. This paper thus introduces an approach relying on the FRAM to methodically derive alarm conditions and system extensions as a foundation for broader uses to investigate in future work. The FRAM-based approach was shown to reduce uncaught faults, required operator interventions, and thus unplanned downtime, in an experiment with two participants on a lab-sized demonstrator machine.

Keywords: Human-Computer Interaction, Multi-User Interaction
Abstract: Real-time collaborative programming environments support a group of programmers, who are geographically distributed, to concurrently view and edit a shared set of source code in a real-time fashion. However, this emerging technology has not been widely applied yet, and one critical challenge is the semantic conflict during real-time collaboration. Existing conflict prevention approaches are limited in supporting individual source code files only, with pre-programmed rules that might be inflexible. To address these challenges, we propose a novel semantic conflict prevention approach based on source code annotations, which allow programmers to manually, conveniently and flexibly apply fine-grained conflict prevention rules during real-time collaboration. The proposed approach and techniques have been successfully implemented in a prototype system. User studies and performance evaluations have demonstrated the technical feasibility of the proposed approach and techniques, as well as the prototype's satisfactory efficiency and usability.

Keywords: Biometric Systems and Bioinformatics, Hybrid Models of Computational Intelligence, Application of Artificial Intelligence
Abstract: Predicting clinical outcomes for patients in the intensive care units (ICUs) is crucial for physicians to assess clinical risk and provide timely and appropriate interventions. Current models often fail to predict multiple clinical outcomes simultaneously, and patient similarity has not been fully utilized to improve prediction accuracy. Additionally, electronic health record (EHR) data in the ICUs often contain observations recorded at irregular time intervals, which existing prediction methods do not effectively model. To address these challenges, we propose the ICU-TGNN model. This model combines time attention-based Transformer and graph neural network (GNN) architectures to leverage temporal and relational patient data. The Transformer component analyzes time series EHR data to capture dynamic patient states over time, while the GNN component exploits the relational structure among patients based on comorbidities to enhance model generalizability. We evaluated the ICU-TGNN model on the eICU-CRD dataset, demonstrating its superior capability in predicting general clinical outcomes such as ICU mortality and length of ICU stay. Our findings highlight ICU-TGNN’s capability to provide accurate outcome predictions by effectively handling the complexity of ICU data, thereby holding great potential to optimize patient management and improve clinical outcomes.

Keywords: Biometric Systems and Bioinformatics, Deep Learning, Application of Artificial Intelligence
Abstract: Molecular conformation generation poses a significant challenge in the field of computational chemistry. Recently, Diffusion Probabilistic Models (DPMs) and Score-Based Generative Models (SGMs) are effectively used due to their capacity for generating accurate conformations far beyond conventional physics-based approaches. However, the discrepancy between training and inference rises a critical problem known as the exposure bias. While this issue has been extensively investigated in DPMs, the existence of exposure bias in SGMs and its effective measurement remain unsolved, which hinders the use of bias compensation methods for SGMs, including ConfGF and Torsional Diffusion as the representatives. In this work, we propose a method for measuring exposure bias in SGMs used for molecular conformation generation, which can confirm the significant existence of exposure bias in these models and measure the value of bias. This enables the development of bias compensation algorithms for DPMs to be adapted into SGM. Experimental results show that by introducing the compensation method Input Perturbation, originally used for training DPMs, into SGM-based molecular conformation models can significantly improve both the accuracy and diversity of the generated conformations. By using the IP-enhanced Torsional Diffusion model, we achieve new state-of-the-art performance on the GEOM-Drugs dataset and are on par on GEOMQM9. We provide the code publicly at https://github.com/jia-975/torsionalDiff-ip.

Keywords: Swarm Intelligence, Deep Learning, Computational Intelligence
Abstract: 联邦学习 （FL） 是一种方法，它允许资源有限的节点进行协作，而无需共享其数据。然而，本地设备的计算性能、数据的非独立性和同分布性（Non-IID）以及有限的通信资源等都不可避免地降低了模型的收敛速度和精度。在Non-IID数据下获得性能优越的联邦学习模型迫在眉睫。在本文中，我们提出了FedCGSU，这是一种基于本地客户端分布中相似不确定性的客户端分组联邦学习方法。FedCGSU通过利用局部分布的相似性来实现分组，从而减少了不同分布的客户之间权重差异的影响。然后，在聚合方法中将损失值和本地数据量相结合，从而减少了计算弱设备对全局收敛速度的影响。利用3个公共数据集进行了实验，结果表明，FedCGSU方法在提高准确性和加快收敛速&#

Keywords: Deep Learning, Neural Networks and their Applications, Application of Artificial Intelligence
Abstract: Personalized News Recommendation (PNR) can enhance user experience by alleviating information overload. Traditional news recommendation methods consider all clicking behaviors as user interests, resulting in biased user modeling that fails to accurately capture user interests. In addition, although there are methods to reduce the impact of low-quality news at the representation level by simply filtering it through the attention mechanism. However, this only implicitly models the news in the interaction sequence without specifically considering the quality of each news, and thus has very limited effect in identifying noise. To address these issues, this paper proposes News Recommendation method based on News Quality-aware modeling (NQNR). We attempt to explicitly model the news in the click sequence and andidate ranking one by one to visually assess the quality of each news. Specifically, we design a detection module to detect whether the input news is low-quality news. Then, by reducing the influence of low-quality news in user modeling and candidate ranking, user interests are modeled more accurately, while recommendations of such news are reduced for users. In addition, to capture the similarity of vectors more accurately, we also design a similarity computation method based on the multiple attention mechanism in the detection module. Experiments on a large real-world Microsoft News Dataset (MIND) show that our model significantly outperforms previous models. Our code is posted at the following URL: https://github.com/xxbbjj/NQNR-.

Keywords: Deep Learning, Neural Networks and their Applications, Machine Learning
Abstract: Bayesian inference has provided the continual learning (CL) with an elegant framework where past experiences and new knowledge are consolidated into the posterior constantly. Typical approaches rely on Bayesian neural networks whose parameters are updated by variational inference, namely, maximizing the evidence lower bound of log-likelihood. In this paper, we discuss the effects of local reparameterization on the optimization of such networks in the context of CL. The empirical results show that it does not only increase the inference speed of neural networks, but also enhance the CL performance in some scenarios. Additionally, motivated by the observation that variance matrices have low-rank structures, we propose the d-tied variational continual learning (d-tied-VCL) to improve the parameter efficiency of variational continual learning (VCL). Experiments on random classification, permuted MNIST, and split CIFAR100 show that even VCL with rank-1 variance matrices achieves competitive performance.

Keywords: Deep Learning, Neural Networks and their Applications, Application of Artificial Intelligence
Abstract: Personalized news recommendation is the process of predicting the relevance of news to users and recommending news to user to fulfill their information needs. However, existing news recommendation methods extract semantic information from users and candidate news respectively, ignoring semantic interaction information between users and candidate news. Furthermore, previous models only use same node types for message passing, ignoring different characteristics and topology between different node types. In addition, existing methods learn news representations through text representations, ignoring semantic correlation information between entity relationships and texts. To solve these problems, we propose a personalized news recommendation model named CoHG. In our model, we design a collaborative fusion module to obtain semantic interaction information through interacting with user history news and candidate news. Then, we design a heterogeneous gated graph neural network that maps different node types into a same space to extract higher-order information in user graphs for message passing. Moreover, we design an enhanced relevant attention module to enhance semantic correlation information of text content by aggregating text representation and entity representation into a unified representation. Finally, we conducted experiments on MIND and Adressa datasets to compare with other baseline models.

Keywords: Neural Networks and their Applications, Deep Learning, Representation Learning
Abstract: With the popularity of location-based services, point of interest (POI) recommendation has become a hot research topic. Current researches mainly focus on the analysis of personal check-in trajectories to obtain user preferences. However, a user’s check-in data is generally sparse, and it is difficult to make accurate recommendation by only using the user’s local information. Additionally, the public’s check-in behavior may exhibit common visiting patterns, and incorporating global check-in information is beneficial for enhancing the learning of individual user preferences. Therefore, we propose a GCN-based model with Fusion of Global and Local information (GFGL) for the next POI recommendation. The model obtains global information such as spatial distance, social relationships, and transition probabilities from all users’ visit trajectories, and utilizes graph convolution network (GCN) for learning of multi-dimensional global information. It considers the asymmetry of transition probability graphs and the symmetry of spatial distance and social relationship graphs, conducting different operations on these two types of graphs. Next, we fuse global information with user local information through the user context information embedding module. Besides, a long short-term memory (LSTM) model is employed to dynamically perceive the transitions of POI categories and the sequential behaviors of users. A transformer model is utilized to explore the relationships between non-adjacent visits in trajectories, understanding the overall preferences of a user. Extensive experiments on two real-world datasets demonstrate the superiority of GFGL against state-of-the-art methods in the next POI recommendation.

Keywords: Deep Learning, Representation Learning, Transfer Learning
Abstract: In Natural Language Processing (NLP), understanding the semantic connection between two texts, a semantic text similarity (STS) task, is a significant challenge. Especially in resource-restricted and cross-domain settings, traditional methods are limited by the high cost of data labeling. This challenge is particularly evident in resource-restricted and cross-domain settings, where traditional methods face high data labeling costs. We propose an innovative method, designated as ``Cold-Start Self-Training" that reduces reliance on large labeled datasets for STS tasks in resource-restricted settings. This method utilizes dual-view pooling to extract semantic similarity information from unlabeled data and generates pseudo-labeled data to fine-tune the cross-encoder model. Dual-view pooling combines different pooling results of the same text to evaluate semantic similarity without additional model tuning, simplifying the self-training process. Experimental results show that our method significantly improves the cross-encoder model's performance on STS tasks in the medical domain. Our findings provide new strategies for cross-domain STS tasks, challenging the traditional reliance on extensive labeled data. We also validate the potential of unsupervised pre-trained models for cross-domain tasks, offering theoretical and practical support for complex challenges.

Keywords: System Architecture
Abstract: In Linux kernel version 6.6, the Earliest Eligible Virtual Deadline First (EEVDF) scheduler was introduced as the new default scheduler. However, due to its high computational complexity, it may not be suitable for all application scenarios. In cases where there are a large number of short-term tasks or frequent task communication, EEVDF can incur excessive context switch overhead and performance degradation. To address these issues, we propose a lightweight scheduling strategy named SRAND. Leveraging the programmable scheduling framework `sched_ext` and employing BPF technology, we have implemented a strategy based on global and local run queues. This strategy utilizes a five-level BPF mapped queue to partition tasks with different virtual runtimes. Tasks with smaller virtual runtimes are placed into a FIFO-type global queue first, enabling priority scheduling for tasks with smaller virtual runtimes. Additionally, we monitor CPU idle states and allocate tasks in a timely manner, enhancing task responsiveness while reducing scheduling complexity.It is worth noting that our strategy integrates user-space scheduling policies into the kernel via an eBPF program loader, thus eliminating the need for kernel code modifications. By implementing the SRAND strategy, we have observed significant improvements compared to Linux's default scheduling strategy EEVDF. Specifically, our proposed strategy averages an 11.83% reduction in process context switch time and an overall performance improvement of 7.02% in stress tests, while maintaining satisfactory load balancing.

Keywords: Robotic Systems, System Modeling and Control, Digital Twin
Abstract: This paper discusses developments for a multi-limb morphogenetic UAV, MorphoGear, that is capable of both aerial flight and ground locomotion. A hybrid path planning algorithm based on the A* strategy has been developed, enabling seamless transition between air-to-ground navigation modes, thereby enhancing robot's mobility in complex environments. Moreover, precise path following is achieved during ground locomotion with a Model Predictive Control (MPC) architecture for its novel walking behaviour. Experimental validation was conducted in the Unity simulation environment utilizing Python scripts to compute control values. The algorithm's performance is validated by the Root Mean Squared Error (RMSE) of 0.91 cm and a maximum error of 1.85 cm, as demonstrated by the results. These developments highlight the adaptability of MorphoGear in navigation through cluttered environments, establishing it as a usable tool in autonomous exploration, both aerial and ground-based.

Keywords: Intelligent Power Grid
Abstract: This paper explores the estimation of the State of Charge (SoC) of lithium-ion batteries. Currently, the majority of research efforts focus on the SoC estimation of individual lithium-ion batteries. However, in practical scenarios, lithium-ion batteries are commonly connected with balancing circuits to address battery imbalances. Upon activation of the equalization circuit, the battery's system dynamics transition to a new mode. Therefore, it is difficult for classical SoC estimation algorithms to accurately estimate the real SoC value. In this paper, we employ a switched system methodology to estimate the battery's SoC. We describe the switched system of the Thevenin equivalent circuit model of a lithium-ion battery using a switched resistance balance circuit. Then we use the method of nonlinear switching observer to analyze the convergence and divergence. Finally, we set up an experimental platform and verify the performance of the observer through several sets of experiments.

Keywords: Intelligent Green Production Systems, Autonomous Vehicle, System Modeling and Control
Abstract: Traffic congestion frequently occurs on slope segments of highways, which is a typical bottleneck. With the development of connected and autonomous vehicle (CAV) technology, there will be a scene that CAVs and human driven vehicles (HDVs) co-exist. This work studies a slope bottleneck on highway in mixed traffic scenarios and proposes a traffic flow model for slope bottlenecks incorporating CAV platooning based on cellular automata. A novel traffic flow control strategy for slope bottlenecks is proposed based on variable speed limit (VSL) and vehicle platooning. Firstly, it divides the upstream section of the slope bottleneck into two zones for implementing VSL and vehicle platooning, respectively. Via speed restrictions within the VSL zone, the inflow of vehicles into the vehicle platooning zone is effectively mitigated to create low traffic density. In the vehicle platooning zone, a hybrid vehicle platooning method for mixed scenarios is proposed. The experimental results demonstrate that our strategy effectively enhances traffic flow of the slope bottleneck, thereby mitigating traffic congestion.

Keywords: Communications
Abstract: The Global Navigation Satellite System Network Reference Stations (GNSS-NRS) are pivotal in modern positioning and navigation applications. However, GNSS-NRS face significant communication challenges, including instability during rapid user movement and inconsistent coverage by different operators, which often results in data transmission interruptions and delays. Moreover, the conventional Multipath Transmission Control Protocol (MPTCP) scheduling strategy does not accommodate the diversity of the data types managed by GNSS-NRS. This default approach treats all data uniformly, even under deteriorating network conditions. To overcome these limitations, this paper introduces a 'Multiple Reception' strategy based on MPTCP that ensures timely and reliable GNSS data transmission. This novel strategy differentiates between observational and control data based on their real-time criticality and applies tailored transmission techniques. Under adverse conditions, it utilizes multiple transmission paths and selectively receives control data, thus prioritizing the transfer of critical information. We conducted simulation experiments using MiniNet to compare the 'Multiple Reception' method against conventional and redundant strategies, demonstrating that it significantly reduces data loss rates in GNSS-NRS.

Keywords: Cooperative Systems and Control, Distributed Intelligent Systems, Modeling of Autonomous Systems
Abstract: Multi-UAV cooperative path planning (MUCPP) is one of the core issues of UAV swarms. When the number of UAVs is large or the mission is complex, problems such as poor scalability and slow convergence will be faced. This paper proposes a multi-UAV distributed hierarchical collaborative path planning method based on an improved particle swarm optimization algorithm. In order to speed up the convergence and ensure the global optimality of the results, a coefficient nonlinear time-varying parallel particle swarm optimization algorithm (NTVPPSO) is designed in this method. This algorithm splits the population into multiple sub-populations. Each particle in the sub-population is updated in parallel using non-linear time-varying coefficients, and information is exchanged between the sub-populations. In order to enhance scalability, a distributed framework with a two-layer structure is designed. First, each UAV plans its own path to determine the collaborative goal. Then the collaborative path planning of multiple UAVs is completed sequentially based on heuristic priorities. The simulation results show that the proposed method can converge quickly and the collaborative path tends to be optimal, and is a feasible distributed solution to the MUCPP problem.

Keywords: Cooperative Systems and Control, Distributed Intelligent Systems, System Modeling and Control
Abstract: This paper addresses the maneuver control problem for the multi-agent cooperative transportation systems (MACTSs) with double-integrator dynamics over switching formations. The switching formations consists of a pair of operations: one is the switching directed graphs and the other is the corresponding formation configurations. In real cooperative transportation scenarios, varying interaction relationships necessitate distinct system configurations. But it is challenging for researchers to design control law with the evolutions of not only the communication topology but also the system configuration. Drawing inspiration from advancements on switching topologies, we utilize the characteristics of the stress matrix to build up a new LMI inequality to design a novel class of distributed affine-based controller. The global convergence will be achieved as long as the feedback gain matrix and the switching signal satisfy three specific conditions. And we give the corresponding algorithm to calculate the required control parameters. A Lyapunov function is constructed to demonstrate the system's stability and simulation example is provided in detail at the end of this paper to validate our method's efficacy.

Keywords: Cooperative Systems and Control, Cyber-physical systems, Modeling of Autonomous Systems
Abstract: This paper proposes the Dynamic and Cooperative Multi-Agent Task Allocation (DC-MATA) problem, focusing on individually rational agents in a cooperative organization, which allocate dynamically changing tasks over time. DC-MATA aims at dynamically improving Nash equilibrium over time through learning in this context. Task utilities evolve dynamically, and learning, conducted in rounds, optimizes agents' task selection order to enhance system performance. Our proposed DC-MATA solution approach assigns agents to tasks with highest utility over time and tends towards the Nash equilibrium that aligns with agents' self-interest while improving the gap with system optimum. We propose priority-sensitive reward function and four action sampling algorithms (varepsilon-greedy, varepsilon-decay, Adapted Simulated Annealing, and Prior Sequence-Aware Sampling - PSAS) leveraging a Markov decision process (MDP) framework. Simulation experiments on our newly proposed GitHub benchmark instances confirm robust performance, facilitating efficient task allocation in the DC-MATA scenario.

Keywords: Biometric Systems and Bioinformatics, Deep Learning
Abstract: If drivers can be correctly recognized from their physiological signals, personalized services such as dynamical adjustments to the seat, backrest, and headrest can be provided during driving to ensure comfort and safety. However, such services require high accuracy identification of drivers, i.e., identification accuracy more than 99%. To determine whether such high accuracy can be achieved using drivers' physiological signals, we propose identifying only a limited group of drivers for a particular vehicle. Specifically, we built two deep learning models from three common physiological signals. To achieve high accuracy identification of drivers, we adjusted the size of the drivers to be identified (driver pool) to achieve a high identification accuracy of drivers. Our results show that when the maximum driver pool sizes are 5 and 2, the identification accuracies reached 95% and 99%, respectively.

Keywords: Complex Network, Machine Learning, Knowledge Acquisition
Abstract: Dynamic community detection is crucial for elucidating the temporal evolution of social structures, information dissemination, and interactive behaviors within complex networks. Nonnegative matrix factorization provides an efficient framework for identifying communities in static networks but fall short in depicting temporal variations in community affiliations. To solve this problem, this paper proposes a Modularity maximization-incorporated Nonnegative Tensor RESCAL Decomposition (MNTD) model for dynamic community detection. This method serves two primary functions: a) Nonnegative tensor RESCAL decomposition extracts latent community structures in different time slots, highlighting the persistence and transformation of communities; and b) Incorporating an initial community structure into the modularity maximization algorithm, facilitating more precise community segmentations. Comparative analysis of real-world datasets shows that the MNTD is superior to state-of-the-art dynamic community detection methods in the accuracy of community detection.

Keywords: Cooperative Systems and Control, Adaptive Systems, Decision Support Systems
Abstract: Most project development processes are iterative and can be divided into multiple tasks. One person can take on multiple tasks, and one task can be assigned to multiple people. The many-to-many personnel allocation method greatly improves the efficiency of the project and saves the cost of the project. There will be two different losses in this allocation plan: the tasks undertaken by personnel are too discrete and the personnel are easily distracted when undertaking important tasks. This paper first formally models the project personnel allocation problem through the Group Multi Role Assignment (GMRA) model. Then two new constraint formulas were proposed to extend the GMRA model to reduce the loss of personnel allocation and the necessary and sufficient conditions of the extended method were proved. Subsequently, two large-scale simulation experiments were carried out to compare and demonstrate the differences between the expanded new method and the original model, and to explore the sufficient and necessary conditions to increase the speed of finding feasible solutions for the new method. Using the improved model for arranging personnel of development projects not only enables efficient many-to-many allocation but also helps reduce a lot of hidden losses in the project process.

Keywords: Decision Support Systems, Cooperative Systems and Control
Abstract: Abstract—The Bank Credit Decision-Making Problem (BCDMP) is one of the main issues that bank operations need to face. To obtain the maximum profit value and optimal loan plan of the bank as much as possible, this article suggests converting BCDMP into a Many-to-Many Assignment Problem, which can be specified by the Multi-Role Assignment (GMRA). GMRA is a sub-model of the E-CARGO. By revised GMRA, the relationship between the enterprises and loans is converted into the relationship between agents and roles, and a multi-dimensional and multi-index evaluation method is used to evaluate the matching degrees between enterprises and loans. We use the Entropy Weight Method (EWM) and the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to obtain the enterprises' score for a loan through four indicators: profits, inventory, turnover ability, and credit. Then we considered the impact of different loan interest rates on bank profits, obtained enterprise scores under different loan interest rates, and used linear programming to solve the problem, achieving good results (The bank achieved a profit margin of 5.43176% via revised GMRA). Keywords—E-CARGO, The Bank Credit Decision-Making Problem (BCDMP), Group Multi-Role Assignment (GMRA), Revised Group Multi-Role Assignment (RGMRA), Entropy Weight Method (EWM), Technique for Order Preference by Similarity to Ideal Solution (TOPSIS).

Keywords: Machine Vision, Application of Artificial Intelligence, Deep Learning
Abstract: 在工业制造中，视觉异常检测是 通过检测和 防止生产异常。异常检测方法 基于知识蒸馏，展现出前途无量 在解决不可预测性方面的绩效，以及 异常的多样性。然而，他们缺乏 在处理时，来自无异常先验的有效指导 异常特征和未充分利用的多尺度特征 在分割评分阶段，产生次优结果 检测结果。为了缓解这些问题，我们提出了一种 无异常先验指导知识蒸馏 （APG） 工业异常检测。首先，它过滤 通过训练去噪目标网络来呈现异常特征 具有知识蒸馏结构。同时，我们 提出先验感知传播模块（P3M）， 通过以下方式提取更有效的无异常特征 对异常特征施加约束。其次，我们 提出多尺度先验引导融合模块（MPGFM） 利用 作为先验目标网络的无异常特征 &#

Keywords: Machine Vision, Application of Artificial Intelligence, Deep Learning
Abstract: 异常检测通常依赖于大量的训练集 样本和异常样本相对罕见且 在日常工业场景中很难获得。在 Few-Shot 异常检测 （FSAD） 的方案，如何 更好地利用少数图像的细粒度特征 是一个关键问题。为了解决这些问题，更细粒度的 本文提出了暹罗网络（FGSNet）。FGSNet公司 考虑 Few-Shot Anomaly Detection （FSAD） 的设置 由两个阶段组成。第一阶段是负责 用于提取更细粒度的特征并大致对齐 该图像的特点，同时结合了高效的 细粒度特征融合模块 （F3M） 以增强特征 表示法。同时，我们设计了一个深度监督损失 提升信息细粒度水平 萃取。第二个主要用于提炼和 对融合的特征进行去噪，然后进行详细的对齐 后续特征分布建模的操作。 与大多数使用单类的现有 FASD 模型相比 训练和单模型评估&

Keywords: Deep Learning, Application of Artificial Intelligence, Representation Learning
Abstract: Industrial time series prediction (ITSP) is an indispensable part of predictive control in modern industry. Recently, supervised deep learning-based methods have provided solutions with sufficient annotated data. However, there is massive unlabeled data with complex temporal features in modern industrial production, resulting in poor performance of these methods. To address this problem, a self-supervised generative pre-trained model with a learnable mask network (SSGPM-LMN) is proposed in this paper. First, the multivariate time series are made into patches channel-independently. Then, these patches are fed into a Transformer encoder with the learnable mask-reconstruction paradigm, drawing mask indices with high temporal features by calculating the cosine similarity in low-dimensional feature space to better learn general representations. Furthermore, a two-step fine-tuning strategy, including linear probing and full fine-tuning, is adopted for various downstream scenarios. Finally, extensive experimental results on case studies of ITSP and transfer learning indicate that our SSGPM-LMN achieves superior performance.

Keywords: Computational Intelligence, Deep Learning, Evolutionary Computation
Abstract: Darwinian evolution of the biological brain is documented through multiple lines of evidence, although the modes of evolutionary changes remain unclear. Drawing inspiration from the evolved neural systems (e.g., visual cortex), deep learning models have demonstrated superior performance in visual tasks, among others. While the success of training deep neural networks has been relying on back-propagation (BP) and its variants to learn representations from data, BP does not incorporate the evolutionary processes that govern biological neural systems. This work proposes a neural network optimization framework based on evolutionary theory. Specifically, BP-trained deep neural networks for visual recognition tasks obtained from the ending epochs are considered the primordial ancestors (initial population). Subsequently, the population evolved with differential evolution. Extensive experiments are carried out to examine the relationships between Darwinian evolution and neural network optimization, including the correspondence between datasets, environment, models, and living species. The empirical results show that the proposed framework has positive impacts on the network, with reduced over-fitting and an order of magnitude lower time complexity compared to BP. Moreover, the experiments show that the proposed framework performs well on deep neural networks and big datasets.

Keywords: Evolutionary Computation, Computational Intelligence
Abstract: The rapid growth of electric vehicles (EVs) has led to significant challenges in providing efficient and sustainable charging solutions. This paper addresses the battery swapping station (BSS) recommendation problem by proposing a novel conflict-free genetic algorithm (CFGA) integrated with a Nash equilibrium seeking (NES) approach to identify optimal Nash equilibrium (ONE) solutions to such a non-cooperative optimization problem. The CFGA employs specialized crossover and mutation operators to generate offspring that satisfy the constraints of the problem, ensuring that each EV decides a unique battery swap strategy without conflict. Firstly, an order crossover operator is proposed to preserve the order of genes in the chromosomes. Secondly, a replacement and exchange mutation operator is proposed to enhance mutation diversity. The resulting optimal solution is then used as the initial strategy for the NES, which iteratively converges to the ONE. The proposed CFGA with NES algorithm is evaluated under both small-scale and large-scale cases, demonstrating its effectiveness in achieving a balance between costs for EVs and utilization for BSSs. The study’s findings have practical implications for the smart grid and EV integration, offering a robust method for optimizing EV infrastructure and operations.

Keywords: Evolutionary Computation, Computational Intelligence, Image Processing and Pattern Recognition
Abstract: Data-efficient image classification poses a challenge in achieving effectiveness with limited data, as evidenced by the current methods based on convolutional neural networks (CNNs) and genetic programming (GP). Existing works employing these two methods encounter limitations, such as a lack of flexibility and an inability to effectively explore the latent features of the data. To tackle these challenges, this paper introduces a genetic programming method for data-efficient image recognition, leveraging novel function sets, terminal sets, and program structures. This method transforms tree-based data structures in GP into trainable CNN architectures. Further, by employing block structures instead of single operations in the search space, the search space is reduced and the stability of the search structures enhanced. Comparative experiments with state-of-the-art neural network methods and GP-based methods on data-efficient classification datasets validate the GP-CNN method offering higher performance.

Keywords: Deep Learning
Abstract: Ellipsis phenomenon is prevalent, particularly in languages like Chinese, and can be observed across various domains, including everyday conversations, literary works, and product evaluations. Such ellipsis poses challenges for machines, affecting the performance of natural language processing tasks such as machine translation and comprehension. Current approaches to tackle ellipsis primarily involve fine-tuning pre-trained language models like BERT. This approach encompasses two subtasks: 1) detecting ellipsis positions, i.e., identifying locations where ellipsis occurs within a sentence, and 2) completing the elliptical content by predicting the missing information based on the detected positions. This paper proposes a novel model that combines fine-tuned BERT with Generative Adversarial Networks (GANs), incorporating a sampler between the generator and discriminator. The proposed method simultaneously utilizes the generator, discriminator, and sampler to locate and complete ellipsis. Experimental results demonstrate that our approach achieves an F1 score improvement of 0.03 in ellipsis position detection and an EM (Exact Match) improvement of 0.1 in ellipsis content completion compared to the baseline method.

Keywords: Deep Learning, Neural Networks and their Applications, Image Processing and Pattern Recognition
Abstract: Various defects arise during textile production, making fabric defect inspection essential for production and quality management in the textile industry. However, fabric defect detection techniques face challenges due to defects with disparate aspect ratios, high foreground-background similarity, and tiny sizes. Based on these issues, we propose a fabric defect detection method that combines an improved Cascade R-CNN (SPCNet) network and Super-Resolution reconstruction technology. Firstly, defective images are reconstructed using a hybrid attention Transformer (HAT), enhancing texture details and edge information. We design a new multi-stage defect detection model SPCNet to identify fabric defects. The architecture includes a feature extraction module based on Switchable Atrous Convolution (SAC). SAC can obtain a larger receptive field for better detection of tiny defects. Path Aggregation Network (PANet) is introduced to improve the recognition of scale-unbalanced defects. In addition, Cascade RPN (C-RPN) is adopted to fully use deep and shallow features. To solve the issue of imbalanced defect classes, we adopt Class-aware Sampling (CAS) strategy. Soft-NMS is used to reduce the false deletion of defective feature detection boxes. Comparative experimental results demonstrate that the defect detection method combining HAT and SPCNet can significantly raise the overall recognition rate of multi-class fabric defects, exceeding the performance of other current methods.

Keywords: Multimedia Computation, Deep Learning
Abstract: Multi-modal physiological signals fusion integrates multiple heterogeneous physiological signals to characterize human physiological activities, which is basic research on biomedical signal processing. Most researches on multi-modal physiological signal fusion are based on heterogeneous graph fusion networks (HGFNs) while they ignore the high-order correlations between multi-modal physiological signals. Moreover, most HGFNs suffer from the over-smoothing problem, causing failures in deeper networks. To address these issues, this paper proposes a deep hypergraph convolutional network called DHFusion for multi-modal physiological signal fusion. Specially, DHFusion designs deep hypergraph convolution (DHGCN) layer to effectively integrate multiple physiological signals and obtain multi-modal features. DHFusion then introduces JK readout layer to enable multi-layer DHGCN to capture deep multi-modal features based on high-order correlations. DHFusion effectively solves the over-smoothing of HGFNs and performs well in representing deep multi-modal features. Experimental results over the state-of-the-art methods on two benchmark datasets demonstrate the effectiveness of the proposed method.

Keywords: Machine Learning, Deep Learning, Complex Network
Abstract: We analyze datasets collected by the Center for Applied Internet Data Analysis (CAIDA) and the Réseaux IP Européens (RIPE) sites to demonstrate the efficacy of machine learning models in predicting Internet outages and disruptions that affect and hinder network access to essential services and humanitarian aid in times of conflict. Custom datasets for specific geographic areas are created using data filtering. A sliding time window transformation was applied to design models that accurately predict disruptions. Model performance is evaluated based on the prediction accuracy using Internet Outage Detection and Analysis (IODA) and RIPE datasets.

Keywords: Machine Learning, Deep Learning, Neural Networks and their Applications
Abstract: At CRYPTO’19, Gohr[1] presented ResNet-based neural distinguishers (ND) for the round-reduced SPECK32/64 cipher. However, due to the black-box use of such deep learning models, it is hard for humans to understand why these distinguishers work, impeding advancements in cryptanalytic knowledge. In this work, we aim to effectively adapt eXplainable Artificial Intelligence (XAI) techniques, notably Local Interpretable Model-Agnostic Explanations (LIME) and Shapley Additive Explanations (SHAP), to gain a detailed understanding of the important features useful in Gohr’s neural distinguishers.

Keywords: Neural Networks and their Applications, Deep Learning, Machine Vision
Abstract: 视觉问答 （VQA） 呈现重要 计算机视觉和计算机视觉交叉点的挑战 自然语言处理。为了更好地理解 问题与图像的关系，我们的模型 引入了一种名为CMACC（Cross-Modal 基于对抗性对比学习的共同注意）。中国中华人民共和国食品协会 通过一个有效地整合了图像和文本信息。 采用跨模态对抗的协同注意力机制 对比学习。对抗性学习使 文本和图像之间的潜在特征分布，而 对比学习使跨 多模态样本描述相同的上下文，从而 增强模型对多模态输入的适应性。 提高鲁棒性，促进更多人的学习 广义特征表示，对比学习 作为损失函数的一个组成部分被合并， 明确强调感知之间的差异 相似和不同的特征。此外，高级 数据增强技术被集成以进一步 增强模型对不同&#

Keywords: Brain-based Information Communications, Systems Safety and Security, Supervisory Control
Abstract: Obstacle avoidance is a crucial task in unmanned aerial vehicles (UAV) motion planning. The accuracy and consistency of real-time visual information affect the generation of obstacle avoidance commands, raising higher safety demands for obstacle avoidance. The neuromorphic computing-based obstacle avoidance solution can address these challenges. Dynamic vision sensors (DVS) exhibit low latency, low power consumption, and high dynamic range as novel neuromorphic sensors. Spiking neural networks (SNN) also leverage the same mechanism to efficiently process asynchronous and sparse event data generated by DVS, offering latency and energy efficiency advantages. Additionally, the optimal estimation method effectively mitigates the impact of noise and interference within the system, reducing the influence of errors on the algorithm and enhancing safety. Based on these considerations, this paper proposes a fast and safe obstacle avoidance framework. DVS is used to acquire event data from the environment, and a hardware-compatible lightweight SNN is employed to extract dynamic obstacle position information from the data. Compared to baseline methods, this approach reduces latency by 85%. Furthermore, two estimation methods are used to predict the movement of obstacles, ensuring flight safety by generating different UAV obstacle avoidance actions based on confidence intervals, even in the presence of obstacle information errors and omissions.

Keywords: Systems Safety and Security,, Systems Safety and Security, Networking and Decision-Making
Abstract: The Internet of Things (IoT) is vulnerable to cybersecurity attacks because it is typically designed with minimal security considerations. The Low-Rate Distributed Denial-of-Service attacks (LR-DDoS) evade detection by injecting a small portion of malicious packets that persistently lurk within the IoT network. Recently, adversarial LR-DDoS attacks by machine learning approaches further complicate the protection. In this study, an improved detection algorithm based on a generative adversarial network (GAN) is proposed against LR-DDoS. To learn the temporal characteristics, LR-DDoS time-series features are taken as a part of the generator’s input and LSTM (Long Short-Term Memory) modules are incorporated. Two discriminators with adaptive weighting are employed, one for conventional LR-DDoS attacks and the other for the adversarial attacks, to generate plausible LR-DDoS samples. The generated packets are used to train the detection algorithm of an Intrusion Detection System (IDS). Simulation results show that the proposed method achieves an accuracy of 94% in detecting adversarial LR-DDoS packets, which outperforms the state-of-the-art methods by at least 10%.

Keywords: Systems Safety and Security,, Systems Safety and Security
Abstract: Federated learning addresses the issues of data silos and privacy to some extent by training models locally on client devices, only uploading model parameters, and aggregating them on the central server. However, attackers can still infer private data information from the uploaded parameters. To solve this issue, differential privacy technology is introduced. The key to differential privacy lies in gradient clipping and noise addition. However, the traditional gradient clipping method often faces the gradient distortion issue and will be ineffective if the noise is large enough. Regarding noise addition, commonly used fixed or fixed decay rate noise scale settings often overlook the characteristics of gradients during training, which might lead to adding improper noise to gradients. To address these issues, we propose an adaptive optimization method for differentially private federated learning (AODPFL). Specifically, we adopt a strategy of clipping the gradient by grouping, effectively reducing gradient distortion. We design an adaptive clipping threshold based on the gradient changes during training to improve model accuracy under large noise conditions. We design a noise scale decay method with a dynamic decay rate to allocate the privacy budget more reasonably and inject appropriate noise into gradients. Experimental results show that compared to other gradient clipping and noise addition methods, our method achieves higher accuracy under the same privacy budget.

Keywords: Systems Safety and Security
Abstract: Nowadays, vision transformers (ViTs) are one of the most prominent state-of-the-art models for vision-based tasks. ViTs are being used widely in many safety-critical applications ranging from health care to automotive. However, these widespread deployments also bring the risk of adversarial attacks on ViTs to the forefront. Thus, understanding the vulnerability of vision transforms against possible adversarial attacks is necessary before deployment in real-time scenarios. Adversarial patch attacks represent a practical threat to the viability of ViT-based real-world applications. This study delves into the vulnerability of vision transformers to such attacks, exploring both single and multi-patch adversarial attacks to gauge the robustness of vision transformers across benchmark datasets, including CIFAR-10, CIFAR-100, Tiny ImageNet, and ImageNet-1k. Experimentally, our findings reveal that poly-multi patch attacks constitute formidable adversarial threats, with vision transformers exhibiting greater vulnerability to poly-multi attacks than single, mono, and split-multi attacks. Additionally, we qualitatively elucidate the impact of patch location on the efficacy of adversarial attacks, providing insights into the factors influencing their effectiveness. Through this study, we aim to enhance our understanding of vision transformers’ susceptibility to adversarial patch attacks, contributing to developing strategies for strengthening their security and resilience in real-world applications.

Keywords: Systems Safety and Security,, Systems Safety and Security, Networking and Decision-Making
Abstract: Data-driven methodologies have exhibited remarkable performance in identifying various cyber threats. However, obtaining well-labeled training samples is enormously expensive and often challenging when tackling practical cyber-security problems, due to the cost and difficulties in data annotation. To address this issue, we propose KI-Mix, a novel pseudo-anomaly generation algorithm for cyber threat detection on the basis of the limited labeled anomalies and a large volume of unlabeled data. In a nutshell, KI-Mix incorporates security domain knowledge into data interpolation to capture more labeled data to facilitate semi-supervised detection on cyber anomalies. We compare the performance of KI-Mix with several commonly applied augmentation techniques, such as Mixup and CutMix to evaluate its effectiveness in limited annotated data settings. Through extensive experiments on five security datasets covering various aspects of network threats, we demonstrate that KI-Mix outperforms other methods with equivalent baseline models. Notably, KI-Mix is model-agnostic to enable any data-driven threats detection models to handle incomplete supervision problems in real-world cyber threat detection.

Keywords: Systems Safety and Security,, Systems Safety and Security
Abstract: Advanced Encryption Standard (AES) is one of the most widely used cipher for data encryption. However, recent studies have pointed out some weaknesses in the subkeys used in AES. Also, the sequential subkeys generation process of AES seems to be inefficient for resource-constrained IoT devices. Therefore, in this paper, we studied how secure and efficient the Secure Hash Algorithm-3 (SHA3) generated subkeys will be for the encryption of IoT data by just replacing the AES subkeys with it. For this purpose we used the statistical test suite provided by the National Institute of Standards and Technology and found that SHA-3 generated subkeys are highly random, and also it is significantly efficient for encrypting the IoT Data. However, as symmetric keys are repeatedly used, therefore, only upto 2 MB data, the encryption is significantly (sim 1300 times) efficient. From the experimental analysis, we find that SEASHA3 generates subkeys that are random, non-linear, and irreversible, making it more secure and efficient than traditional AES subkey generation.

Keywords: Active BMIs
Abstract: Brain-computer interface (BCI) technology enables communication between humans and devices by reflecting users' status and intentions. Electroencephalography (EEG) signals are utilized to capture brain electrical activity with no surgical operation. When conducting motor imagery (MI), one of the endogenous BCI paradigms, the users imagine the movement of muscles used when performing a certain movement without actual physical movement. However, not all subjects show outstanding classification performance in decoding MI-based EEG signals. We propose the novel method that utilizes the weights of the pre-trained model to generate the personalized weights, effectively combining the general MI features with the personalized features. We used the 5-fold cross-validation for evaluating the performances, and conducted the experiments in 3 different pre-trained models (Top--3, Top--5, and Top--7). We compared the performances of our proposed method using the baseline and the full fine-tuning. In comparison to our proposed method with the baseline, our proposed method achieved the improvement of the average accuracies in all pre-trained models, and those values were 0.123, 0.138, and 0.143, respectively. When comparing our proposed method with the full fine-tuning, the average accuracies of our proposed method were the highest in all pre-trained models, and the differences in the average accuracies were 0.012, 0.019, and 0.009, respectively. Hence, we demonstrated the possibility of improving the precision and effectiveness of the EEG-based systems by reflecting the individual differences in EEG signals among the subjects with low classification accuracy.

Keywords: Brain-Computer Interfaces, Brain-based Information Communications, Human-Machine Interface
Abstract: Direct speech brain-computer interface (DS-BCI) technology facilitates mind-reading through brain signals, fostering direct communication and control mechanisms. The applications of DS-BCI are extensive, transcending language communication to encompass fields like healthcare and assistive technologies. Given that Mandarin is the predominant tonal language globally, investigation of the tonal characteristics of this language is imperative for DS-BCI advancement. This study collected electroencephalogram (EEG) data from subjects articulating Mandarin sentences, which were then synchronized with corresponding audio recordings. Each word was segmented at the vocalization onset, with a capture window extending 300 milliseconds prior and 200 milliseconds post-articulation. We further engineered a multi-layered CNN architecture for EEG tone recognition. The rational asymmetry (RASM) feature extraction technique proved instrumental in reducing data dimensionality and enhancing model training, culminating in optimal accuracy to identify Tones 1 through 4 at 25.55%, 30.42%, 35.13%, and 27.04%, respectively. This research advances the field by employing complete sentences for EEG-based Mandarin tone recognition.

Keywords: Passive BMIs, Other Neurotechnology and Brain-Related Topics
Abstract: Many users of hearing aids report challenges when listening to music. In the future, it may be possible to develop hearing aids that have electrodes which monitors brain activity in real-time and adapts the filters on the hearing aid to match the volitions of the user. In music, this could mean amplifying the sound of the instrument the listener wants to hear. One of the first steps in this research is to determine if a machine learning algorithm can identify to which instrument an individual is listening based only on a brief EEG signal. To test this possibility, participants were presented with a series of brief tones that varied in timbre (Trombone, Clarinet, Cello, Piano and Pure Tone) while their ongoing EEG was recorded from 73 electrodes. Linear Discriminant Analysis (LDA) was used. We investigated four different sets of features – Raw EEG, ERP-based features, Harmonics-based features and Regularity-based features. The Raw EEG based classifier performed significantly above chance (37%) when attempting to distinguish between responses to different musical instruments for 5-way classification. More advanced classification algorithms or different features may be able to better distinguish between tones with a musical timbre.

Keywords: Passive BMIs, BMI Emerging Applications, Other Neurotechnology and Brain-Related Topics
Abstract: Inner speech recognition is a modern advancement in Brain computer interfaces (BCI) that facilitates a communication between the computer and the brain in a direct way. It is particularly beneficial for individuals who face communication issues due to speech disability. In this research, an end-to-end brain-computer interface system was developed that uses Electroencephalography (EEG) signals to recognize inner speech straight from the brain. The study utilized an open-access dataset presented on four Spanish words from ten participants. The proposed system involved preprocessing to clean and enhance the signal. Common Spatial Pattern (CSP) was utilized to extract features to improve the discrimination between the selected four words. A Convolutional Neural Networks (CNNs) deep learning model was used to enhance the inner speech recognition performance from spatial features. The result shows that the model can decode the inner speech with an average accuracy of 81.40% for the un-seeing dataset and 89.80% for the entire dataset, indicating that the proposed method outperforms previous works that used the same dataset.

Keywords: Brain-Computer Interfaces, Human-Computer Interaction, Assistive Technology
Abstract: Brain-computer interfaces (BCIs) based on motor imagery (imagined movements, IM) are among the most common BCIs for the rehabilitation of paralyzed patients. However, it is possible that attempted movements (AM) would be more an effective alternative for IM. Unlike IM, AM are difficult to study outside of clinical practice. Nikulin et al. (2008) suggest that quasi-movements (QM) could help model AM in healthy participants without immobilizing interventions. QM result from the amplitude reduction of an overt movement, which leads to the practical absence of electromyography (EMG) response. The performance of QM may have features that may distance QM from AM. Here, we examined the compatibility of QM with a saccade task, which modelled visual interaction with the outside world during the practical use of a BCI. In a study involving 24 volunteers, we used electroencephalography (EEG), EMG, and conducted an extensive survey of the participants. We expected that, compared to IM, QM in the dual-task condition would be easier and less tiring and would be accompanied by greater event-related desynchronization (ERD) of the sensorimotor rhythms. Our hypotheses were based on the assumption that like AM and unlike IM, QM is a more external task, and so is more compatible with the saccade task. We reproduced the effect of greater ERD for QM in the dual-task condition but did not find any significant difference between the difficulty or tediousness of QM and IM. Nevertheless, the survey data gave us important insights into the challenges participants faced when performing QM. Despite EMG values similar to IM, the feeling of muscle tension experienced by the participants correlated with mean EMG values. The main challenge in performing QM by the participants was to make movements without an amplitude. Performing QM conflicted with the illusion of movement that was supposed to accompany them: without proprioceptive feedback, participants doubt the reality of QM. Our results can be used to improve the procedure of QM training, which should bring them closer to genuine attempts of movements in the eyes of participants.

Keywords: Autonomous Vehicle, Robotic Systems, Consumer and Industrial Applications
Abstract: This paper presents an autonomous inspection system that utilizes an octocopter UAV to enhance the structural safety assessments of challenging environments such as bridges, high-rise buildings, and dams. The system features an octocopter equipped with a rotating 3D LiDAR for comprehensive environmental data collection, a three-axis gimbal-mounted camera for high-resolution imaging, and an onboard computer for data processing and control. Our algorithms are designed to recognize environmental structures, identify planar surfaces, and calculate optimal gimbal poses for detailed inspections. The UAV autonomously navigates and performs these tasks, ensuring precise structural assessments. We validate the effectiveness of this system through experiments with the octocopter UAV.

Keywords: Robotic Systems, Fault Monitoring and Diagnosis
Abstract: Networked devices, especially those comprising multiple identical devices, are extensively utilized in industrial scenarios. However, their complexity poses unique challenges in diagnostic processes, demanding efficient methodologies to identify and assess risks. The application of Failure Mode and Effects Analysis (FMEA) for analyzing complex systems, especially those consisting of networked devices, appears to be limited, particularly in identifying critical risk factors. In this paper, we propose a novel pipeline to diagnose networked systems by fusing FMEA with a Delphi (expert) Study. Our approach leverages the collective knowledge of a group of experts through a structured Delphi Study enabling them to contribute individually but also to interact. We demonstrate the applicability of our approach through a case study involving a system of networked mobile robots. Based on a Fault Tree Analysis (FTA), Risk Priority Numbers (RPN) of minimal fault tree cut sets are calculated to identify the most critical mechatronic failures. Our findings show that our methodology provides an RPN ranking that closely aligns with expert insights, highlighting its efficacy in accurately assessing risk in complex networked systems.

Keywords: Autonomous Vehicle, Intelligent Transportation Systems
Abstract: Urban autonomous driving has the potential to enhance both safety and efficiency of transportation in environments also in complex traffic conditions. However, new services and approaches are necessary to manage Autonomous Vehicles in the real traffic. This paper introduces a novel approach to optimize routing in the urban settings by Deep Reinforcement Learning (DRL) techniques. A modular DRL architecture is proposed to obtain a route able to minimize the length of the paths, minimize the number of turns during the travel and select the dedicated lanes. The proposed DRL is implemented on a case study where the agents are trained in a simulation environment for the city center of Bari, a town of Southern Italy.

Keywords: Enterprise Information Systems, Service Systems and Organizations
Abstract: The need for document exchange between people, companies and government increases every day. Consequently, safeguarding documents against potential attackers becomes increasingly crucial. Several attacks have been reported over the past years and the risk of document leak is more present nowadays. To prevent data violation, we need tools to determine the sensitivity degree of documents which allows us to guarantee that only authorized people have access to them and to adapt strategies to sensitivity levels. To achieve this, deep learning techniques have shown good performances in document classification and therefore in sensitivity identification. Such approaches require sufficiently large resources to learn robust models. However, due to the sensitive nature of documents, public datasets are missing to conduct research in this context. In this paper, we experiment with Large Language Models (LLM) to generate a multi-domain dataset of business documents. Utilizing a two-step generation process, we employ several prompting strategies across six language models to create a first dataset of documents classified into 4 sensitivity classes: Public, Internal, Confidential and Restricted. We then relied on human experts to review validate the annotations generated in a sample of documents. Their insights were instrumental in generating the final dataset by identifying the most effective prompting strategies and the top-performing LLMs in both English and French languages for sensitive documents generation. The generated dataset has been tested over two robust baselines.

Keywords: Decision Support Systems
Abstract: The production of automobile engines primarily concerns three workshops: casting, machining and assembly. Any stagnation in any of these stages will have a detrimental impact on the operation of downstream production, resulting in increased costs and production delays. Hence, it is imperative to establish a rational scheduling system for the whole production. Therefore, this study proposes a collaborative scheduling optimization method for multi-level hybrid flow shop of automobile engine. Initially, the collaborative production relationships between the workshops are examined. Subsequently, an optimization model of collaborative scheduling for the multi-stage hybrid flow shop is formulated, focusing on reducing collaborative production costs and minimizing the maximum completion time. Furthermore, considering the discrete nature of the scheduling problem, the study explores encoding/decoding strategies based on random key techniques to bridge the gap between continuous algorithms and discrete problems. Consequently, a distributed reference vector guided evolutionary algorithm (DRVEA) is introduced to solve the model. Finally, the effectiveness and superiority of the proposed method are validated through a case study using practical data from an automobile engine enterprise.

Keywords: System Modeling and Control, Decision Support Systems, Cooperative Systems and Control
Abstract: Premature convergence is a critical issue of Particle Swarm Optimization (PSO). The weak global search capability causes particles trapped in local minima at early stage of learning process. There is several research dedicate to solve this problem over the decade. This paper proposes a new algorithm combined Isolation Forest and Particle Swarm Optimization called Isolation Forest Backward Particle Swarm Optimization (IFB-PSO). The proposed new learning scheme helps particles escaping from local minima. The particle will jump backward to the targeted position when the particle trapped over specific iterations. The destination is precisely selected by Isolation Forest to endow the backward particle hopeful future. IFB-PSO is evaluated by a classic benchmark suite, cart-pole problem, and mountain car problem. Experimental results show that IFB-PSO gets competitive results on the benchmark suite with different dimensions and two control problems in comparison with 11 well-known optimization algorithms. The behavior of backward particles is also analyzed to inspect the utility and efficiency of the backward process.

Keywords: Cooperative Systems and Control, System Modeling and Control, Autonomous Vehicle
Abstract: This study examines how to properly control each autonomous machine at a construction site where human-driven construction machines, an autonomous excavator, and an autonomous truck are coexisting. We proposed to utilize Model Predictive Control (MPC) to control two different control targets, an excavator and a truck, using the same method in a unified manner. By taking advantage of the MPC's ability to handle constraint conditions explicitly, we proposed a control method that allows each moving object to operate safely without contact. In order to control each autonomous machine with MPC, the future behaviors of machines other than itself are needed. In this study, a simple but effective model was used to make predictions for human-operated construction machinery. On the other hand, for autonomous machines, we proposed a method to achieve efficient cooperative behavior by utilizing the calculation results of the MPC of other machine. The effectiveness of the proposed method was confirmed by numerical simulations.

Keywords: Cooperative Systems and Control
Abstract: To ensure optimal functionality of the supercapacitor management system in practical applications, the accurate and robust state of charge (SOC) estimation is crucial, particularly to account for aging effects and varying operating conditions. This paper proposes a switched system-based approach for the co-estimation of SOC and parameters of supercapacitors coupled with balancing resistor circuits. A switched model incorporating an equivalent circuit model is developed to accommodate the activation of equalization within a series-connected supercapacitor pack. The method combines a modified recursive least squares (RLS) algorithm with a switching sliding mode observer (SMO) for real-time parameter adaptation and SOC estimation. The experimental verification under the multi-balancing charging scenarios demonstrates significant enhancements in accuracy and robustness compared to traditional methods employing fixed model configurations and parameters.

Keywords: Cooperative Systems and Control, Cyber-physical systems, Modeling of Autonomous Systems
Abstract: In this paper, we propose and formulate the static and time-extended one-on-one multi-agent task allocation problem with implements and limited autonomy (MATAILA). The objective is to assign tasks to a team of agents statically or over a receding time horizon, while minimizing the overall multi-agent team's cost of performing the tasks and the penalty cost for unaccomplished tasks, all while maintaining sufficient battery level across the team. The basis of the studied problem is the (static one-on-one) axial 3-index assignment problem with the extensions on the time horizon and agents' autonomy. Time-extended MATAILA is a computationally expensive problem, that we simplify by a static MATAILA which focuses only at the tasks pending in the present period and is myopic towards the tasks appearing in the future. We compare the performance of the proposed models in scenarios where all tasks are known a priori. We analyze the performance and scalability of the two approaches experimentally in simulations and show their efficiency in dynamically changing scenarios.

Keywords: Cooperative Systems and Control
Abstract: This work considers the problem of intercepting rogue drones targeting sensitive critical infrastructure facilities. While current interception technologies focus mainly on the jamming/spoofing tasks, the challenges of effectively locating and tracking rogue drones have not received adequate attention. Solving this problem and integrating with recently proposed interception techniques will enable a holistic system that can reliably detect, track, and neutralize rogue drones. Specifically, this work considers a team of pursuer UAVs that can search, detect, and track multiple rogue drones over a sensitive facility. The joint search and track problem is addressed through a novel multiagent reinforcement learning scheme to optimize the agent mobility control actions that maximize the number of rogue drones detected and tracked. The performance of the proposed system is investigated under realistic settings through extensive simulation experiments with varying number of agents demonstrating both its performance and scalability.

Keywords: Cooperative Systems and Control
Abstract: The phenomenon of "involution" and "lying flat" in today's society vividly portrays the social pressure and development difficulties faced by the younger generation. Today, more and more college students are competing for a limited amount of employment resources, which will lead to an increasingly serious phenomenon of involution. However, due to involution, the respective final benefits did not increase significantly. It will bring a lot of negative effects. As involution intensifies, how to select the most suitable employees and optimize team performance has become a major challenge for enterprises. This paper formalizes the problem by extending the Group Role Assignment（GRA) model. Subsequently, we conducted a large-scale social simulation experiment. Some weights and thresholds are introduced to explore what evaluation standards companies should set and make personal potential value and the performance of management trainees reasonable. In this case, the impact of the individual's final assessment score is the most balanced, thereby achieving optimal group performance. Through this method, companies can clarify assessment standards, recruit the most suitable employees, and avoid graduate involution.