Keywords: Deep Learning, Computational Intelligence, Machine Learning
Abstract: Understanding how visual information is encoded in electrocorticography (ECoG) signals is essential for developing accurate and interpretable decoding models. In this study, we propose two novel approaches for multi-class visual classification based on ECoG data recorded from the inferior temporal cortex of macaque monkeys. The first model, MST-ECoGNet, combines traditional signal processing with neural networks by employing the Modified Stockwell Transform (MST) to map ECoG signals into a structured time-frequency-spatial domain. The second model, BiBand-3DECoGNet, replaces MST with a learnable convolutional module and utilizes a 3D spatial encoder to exploit the electrode array structure. Experimental results show that our models significantly outperform prior work, achieving up to a 12.87 percentage point improvement in classification accuracy while reducing model size by a factor of ten and increasing training speed by sixfold. Analysis of feature dimensions reveals that spatial and low-frequency components carry the most relevant information for visual decoding. These findings provide a foundation for further exploration of the neural mechanisms underlying visual object representation in the brain.

Keywords: Deep Learning, Evolutionary Computation, Transfer Learning
Abstract: Procedural level generation in video games has made significant strides, yet achieving high-quality automated level design remains a major challenge. Over the years, techniques have evolved from simple constructive algorithms to advanced Artificial Intelligence (AI) models like Generative Adversarial Networks and Large Language Models. However, the lack of a standardised evaluation framework has hindered direct numerical comparisons and the ability to gauge true progress in the field. To address this gap, we propose an evaluation methodology to benchmark key generation techniques and explore the potential of general-purpose AI models. As a case study, we present a Super Mario Bros level generator powered by ChatGPT, leveraging general-purpose natural language for design tasks. The results show the different strengths and weaknesses of existing models, indicating that traditional algorithms still outperform the most advanced AI methods in this domain, highlighting the need for further innovation to bridge the gap.

Keywords: Deep Learning, Hybrid Models of Neural Networks, Fuzzy Systems, and Evolutionary Computing, Machine Learning
Abstract: In order to address missing-value prediction in sparse datasets, we present VAEGAIN, a novel imputation framework modeled after transformer architectures that combines the adversarial robustness of Generative Adversarial Networks (GANs) with the representational power of Variational Autoencoders (VAEs). To capture intricate feature dependencies and maintain statistical coherence, each linear transformation in the GAN discriminator and encoder–decoder pathways is parameterized.

Because high missingness can skew underlying feature correlations and impair downstream model performance, sparse data poses serious imputation challenges in many real-world scenarios. In order to overcome these problems, VAEGAIN uses three dynamically generated matrices—data, mask, and stochastic noise—to direct an end-to-end, context-aware imputation procedure. In order to ensure accurate imputation and strong feature consistency, the framework is optimized using a composite loss that combines reconstruction error with adversarial objectives.

An experimental evaluation shows that, at a fixed 20% missing rate, VAEGAIN achieves 12-18% higher imputation accuracy than state-of-the-art baselines. Our approach offers a trustworthy and interpretable solution for incomplete data reconstruction in high-stakes domains like healthcare and finance by providing trustworthy probabilistic estimates in addition to imputed values. By establishing a new benchmark for hybrid, uncertainty-aware imputation, VAEGAIN opens the door for probabilistic reasoning on sparse, structured datasets in the future.

Keywords: Deep Learning, Image Processing and Pattern Recognition
Abstract: 为了解决低分辨 9575;、小目标的挑৏ 2; 特征丢失和复杂 5299;剖学的干扰 检测不完全根管 0805;填的结构 在牙科根尖周 X 光片中，本文提 0986;了一种 改进的 YOLOv8 模型。首先，我 0204;设计一个 Convolution 模块，其中 Space-to-Depth 转换 （SDT-Conv） 通过空间深度保 0041;特征图分辨率 可分离的卷积， 6377;效缓解信息 下采样作导致小 0446;标丢失。 其次，我们构建
9968;个动态迭代令୮ 0;聚合器 （迪塔） 架构，可增强全 3616;功能 通过超令牌空间 2858;合进行表示，ߣ 7;及 语义关联，同时 7319;用空间语义 双流注意力机制 4378;化多尺度 功能融合能力， 0174;而提供更丰富௚ 0; 整个网络的特征 0449;息。最后，我ߤ 4; 嵌入一个高效的 多尺度 注意 （EMA） 动态校准机制 检测头，通过以
9979;方式优化特征ࡺ 9;应 跨通道权重自适 4212;，使模型能ࣩ

Keywords: Deep Learning, Image Processing and Pattern Recognition
Abstract: Multi-organ segmentation is essential in clinical applications but remains challenging due to diverse organ morphologies, image noise that degrades local features, and data scarcity limiting the representation of small organs. These challenges often cause poor boundary delineation and loss of fine microstructural details. Previous methods typically struggle to effectively balance global contextual information and fine local features, resulting in suboptimal segmentation performance, especially around complex organ boundaries. To address these limitations, we propose RSTnet, a hybrid CNN-Transformer architecture designed to explicitly capture and fuse multi-scale global and local information. RSTnet uses a dual-encoder structure: one path employs a Swin Transformer backbone enhanced with Convolutional Block Attention Modules (CBAM) to extract robust global context, while the other path is a dedicated Convolutional Neural Network (CNN) encoder that focuses on fine-grained local feature extraction. Rather than integrating the parallel encoder outputs directly, Residual Channel Attention Skip Connections (RCASC) are introduced as attention-enhanced skip connections between encoder and decoder layers to improve feature fusion during decoding. Experimental results on the Synapse multi-organ CT dataset demonstrate that RSTnet outperforms state-of-the-art methods, achieving an average Dice Similarity Coefficient (DSC) of 76.25% and an average Hausdorff Distance (HD95) of 25.24%. These results confirm that RSTnet effectively overcomes the limitations of prior methods, particularly in segmenting complex organ boundaries and capturing microstructural details under challenging conditions.

Keywords: Machine Learning, Intelligent Internet Systems, Deep Learning
Abstract: This paper presents an approach to matrix factorization-based recommender systems that incorporates parameter tuning and noise modeling to improve robustness and generalization. The validation performance is best for the dimensions of the latent features k = 2 with a mean squared error (MSE) of validation loss of 0.464. The results show the impact of Gaussian noise on performance and demonstrate how implicit feedback integration can improve the accuracy of recommendations. They offer practical insights into building scalable and robust recommendation systems

Keywords: Application of Artificial Intelligence, Deep Learning, Machine Learning
Abstract: Sensitive content detection in financial reimbursement documents is fundamental to corporate compliance management. Traditional keyword-based approaches suffer from high false-positive rates due to their inability to capture contextual semantics. We propose TripletAudit, a context-aware framework for sensitive content detection that leverages triplet learning, incorporating a hierarchical architecture combining BERT for semantic embedding, Bi-LSTM for sequential modeling, and multi-head attention for context understanding. Our triplet learning mechanism effectively distinguishes between sensitive and non-sensitive contexts through anchor-positive-negative samples. Experiments on real-world reimbursement datasets demonstrate that TripletAudit achieves state-of-the-art performance with 94.03% accuracy and 93.24% F1-score, significantly outperforming baseline methods in financial compliance scenarios.

Keywords: Information Assurance and Intelligence, AI and Applications, Application of Artificial Intelligence
Abstract: Detecting machine-generated text (MGT) from contemporary Large Language Models (LLMs) is increasingly crucial amid risks like disinformation and threats to academic integrity. Existing zero-shot detection paradigms, despite their practicality, often exhibit significant deficiencies. Key challenges include: (1) superficial analyses focused on limited textual attributes, and (2) a lack of investigation into consistency across linguistic dimensions such as style, semantics, and logic. To address these challenges, we introduce the Collaborative Adversarial Multi-agent Framework (CAMF), a novel archi- tecture using multiple LLM-based agents. CAMF employs specialized agents in a synergistic three-phase process: Multi- dimensional Linguistic Feature Extraction, Adversarial Consis- tency Probing, and Synthesized Judgment Aggregation. This structured collaborative-adversarial process enables a deep analysis of subtle, cross-dimensional textual incongruities in- dicative of non-human origin. Empirical evaluations demon- strate CAMF’s significant superiority over state-of-the-art zero- shot MGT detection techniques.

Keywords: Application of Artificial Intelligence, Deep Learning, Machine Learning
Abstract: With the rapid development of Internet technology, social networks have become essential platforms for communication, and in the process, they generate vast amounts of user data that reflect interests, relationships, and habits. User role classification in social networks is critical for personalized recommendations, precision marketing, and security. In this work, we investigate Graph Neural Networks (GNNs) for user role classification in social networks, and a novel GNN architecture, termed Graph Topology-aware Refinement Network (GTRNet), is designed. GTRNet comprises two modules: Graph Topology Encoding (GTE) and Node Representation Refinement (NRR). They combine the node features and network topological information via convolutional layers to enhance the use role classification performance. The experimental results demonstrate that GTRNet usually outperforms the state-of-the-art methods on the Facebook, Cora, and Citeseer datasets, with micro-F1 score improvements of 0.021, 0.009, and 0.014, respectively. It verifies GTRNet's effectiveness in addressing social network user role classification task.

Keywords: Intelligent Internet Systems, Machine Learning, Application of Artificial Intelligence
Abstract: Significant growth of internet users signifies that more autonomous systems need to be designed. Software-Defined Network (SDN) replaces traditional networks due to their solubility and capability to integrate with artificial intelligence. It also makes networking future poof. A technological shift often comes with shortcomings, and SDN is no exception. Anomaly detection is a crucial task. Existing anomaly detection methods use traditional statistical and machine learning-based models that may perform poorly due to two key issues: (i) the unavailability of all training network-packets at the beginning since these packets are received over time and the training dataset gradually grows over time, and (ii) limited memory and computation capacity of the computing system for model building, preventing the storage of all network-packets that arrive over time for a prolonged period. To address the issues, this paper introduces a novel functionality in SDN called Adaptive anomaly Detection Method (ADM), which utilises an existing technique for facilitating incremental learning so that ADM can handle new anomaly types by keeping track of historical information. ADM is evaluated using four criteria by comparing its performances against four state-of-the-art methods on two publicly available datasets. Our experimental results demonstrate that ADM performs significantly better than existing techniques. The source code of this work has been made publicly available at https://github.com/eferdous/adm

Keywords: Application of Artificial Intelligence, Deep Learning, Machine Learning
Abstract: Just-in-time Software Defect Prediction (JIT-SDP) aims to detect potential defects early, helping to prevent risky code from entering the repository during development. This study evaluates JIT-SDP using pre-trained language models in different architectures and settings. It compares open-source fine-tuned models such as CodeT5+ and UniXCoder with closed LLMs such as GPT and Gemini. This is the first known study to compare trainable open and prompt-based closed decoder-only models for JIT-SDP. The main results show that fine-tuned open models outperform closed models in zero-shot and few-shot scenarios without advanced prompt engineering techniques, and in cross-project tasks, CodeT5+ and UniXCoder surpass previous state-of-the-art results. The findings underscore the value of model architecture, fine-tuning, and expert features for effective defect prediction. Finally, we introduce CodeFlowLM – to our knowledge, the first framework for continual JIT-SDP using pre-trained language models.

Keywords: Application of Artificial Intelligence, Deep Learning, Machine Vision
Abstract: Some traditional action recognition methods are limited to using appearance features to identify human behaviors. However, in real-world environments, human activities often occur in complex and changeable scenes. Therefore, learning the action features in videos from complex environments is the key to action recognition. When performing action recognition tasks in complex spatio-temporal backgrounds, fusing action information from different granularity levels is often a necessary step. In response to this idea, we propose a Multi-Granularity Action Network (MGAN), considering the modeling and fusion of spatio-temporal information with regard to action granularity. The central component of this network is the Multi-Granularity Action Excitation Module (MGAE), which excites features in parallel along different channel groups through different granularities and is capable of modeling multi-scale spatio-temporal information. To enhance the motion information between frames, we propose a Local Motion Module (LMM) for extracting fine-grained motion features. We insert the MGAE and the LMM into ResNet-50 to form a simple yet effective Multi-Granularity Action Network. We have conducted a large number of experiments on three widely used datasets, namely Something-Something V1, Something-Something V2, and UCF-101. The experiments demonstrate that with only a small number of additional parameters and computational cost introduced, the proposed MGAN achieves competitive performance.

Keywords: Robotic Systems
Abstract: Large language models (LLMs) have shown promise in empowering robotics, but their widespread real-world application remains challenging due to two main issues: (1) existing research is “out-of-the-box”, struggling to generalize to new robots and tasks, especially long-horizon tasks, and (2) deploying more general and powerful LLMs exceeds the capabilities of commodity hardware. To address these challenges, we propose the edge-cloud collaborative framework, i.e., SkyNet. We deploy LLMs in the cloud to create initial plans, select executable skills from a predefined library, and send them to the edge-based robot. The robot integrates multiple modules to form a policy network to complete the skills and update the feedback history. Based on the feedback history, the cloud LLMs determine whether to replan. The edge-cloud collaborative approach alleviates the pressure of deploying LLMs on commodity hardware, while the modular design enables easy extension to different tasks or robots without reconfiguring everything. To address the lack of standardized real-world experimental setups, we set up two easily replicable long-horizon tasks on a mobile robot equipped with commodity hardware, analyze the performance of various modules, and demonstrated the effectiveness of SkyNet.

Keywords: Robotic Systems
Abstract: Basketball shooting for humanoid robots is a highly complex and challenging task that requires efficient perception, decision-making capabilities, and precise motion control. Traditional control approaches to humanoid robot basketball shooting often rely on extensive manual coding, which results in high development costs, poor adaptability, and limited capability to handle the complex variations of dynamic environments. To address these challenges, this paper divides the basketball shooting process into three sub-tasks: approach the ball, pick up the ball, and shoot the ball. We address the challenges of perception, decision-making, and control by introducing an attention mechanism in the perception module, proposing a hybrid prioritized experience replay method for the Soft Actor-Critic (SAC) algorithm in the decision-making module, and designing low-level actions for robot control. Through end-to-end training, the robot sequentially completes the three sub-tasks and masters the basketball shooting skill through autonomous learning. The experimental results demonstrate that the proposed method significantly outperforms traditional deep reinforcement learning algorithms in terms of convergence speed, and shooting accuracy.

Keywords: Robotic Systems
Abstract: Multi-pursuer multi-evader (MPME) pursuit-evasion (PE) scenarios are common in real-world applications but pose significant computational challenges as the number of agents grows and dynamics become more complex. Hierarchical decomposition simplifies the problem by splitting it into high-level task assignment (TA) and low-level execution. TA, an NP-hard problem with an exponentially expanding solution space, is addressed here using a novel Reinforcement Learning (RL) approach with both static and dynamic TA in a 2D MPME reach-avoid game. The RL method matches the performance of a combinatorial optimization approach with homogeneous pursuers and outperforms it with heterogeneous ones, demonstrating its effectiveness in complex settings.

Keywords: Robotic Systems
Abstract: Despite the potential the ability to identify granular materials creates for applications such as robotic cooking or earth moving, granular material identification remains a challenging area, existing methods mostly relying on shaking the materials in closed containers. This work presents an interactive material identification framework that enables robots to identify a wide range of granular materials using only force-torque measurements. Unlike prior works, the proposed approach uses direct interaction with the materials. The approach is evaluated through experiments with a real-world dataset comprising 11 granular materials, which we also make publicly available. Results show that our method can identify a wide range of granular materials with near-perfect accuracy while relying solely on force measurements obtained from direct interaction. %Additionally, we performed a case study on granular material transportation, which shows that manipulation skills targeted for a particular material may fail with another one, highlighting the potential benefits of reliable material identification. Further, our comprehensive data analysis and experiments show that a high-performance feature space must combine features related to the force signal's time-domain dynamics and frequency spectrum. We account for this by proposing a combination of the raw signal and its high-frequency magnitude histogram as the suggested feature space representation. We show that the proposed feature space outperforms baselines by a significant margin. The code and data set are available at: https://irobotics.aalto.fi/identify_granular/.

Keywords: Robotic Systems, Autonomous Vehicle
Abstract: Rapid urbanization trends have resulted in a significant increase in global urban population over the past few decades. This growth is projected to continue, further exacerbating traffic congestion in road networks. Urban air mobility presents a promising solution to alleviate this congestion, offering substantial environmental, economic, and societal benefits. However, to fully realize these benefits, autonomous flight and automation are crucial for safely managing the high density of aircraft expected in urban airspaces.

To address this challenge, this work proposes an innovative model predictive control (MPC) framework for generating on-demand, safe 3D trajectories for fleets of unmanned aerial vehicles (UAVs) operating in dense urban environments. Specifically, it introduces a near real-time MPC method designed to enhance the airspace’s capacity to accommodate an increasing number of flights within a confined area, while consistently adhering to stringent safety standards. An extensive simulation study is conducted to demonstrate the effectiveness of the proposed framework in planning safe, on-demand trajectories in complex urban settings.

Keywords: Robotic Systems, Autonomous Vehicle, Adaptive Systems
Abstract: Quadcopters are versatile aerial robots gaining popularity in numerous critical applications. However, their operational effectiveness is constrained by limited battery life and restricted flight range. To address these challenges, autonomous drone landing on stationary or mobile charging and battery-swapping stations has become an essential capability. In this study, we present NMPC-Lander, a novel control architecture that integrates Nonlinear Model Predictive Control (NMPC) with Control Barrier Functions (CBF) to achieve precise and safe autonomous landing on both static and dynamic platforms. Our approach employs NMPC for accurate trajectory tracking and landing, while simultaneously incorporating CBF to ensure collision avoidance with static obstacles. Experimental evaluations on the real hardware demonstrate high precision in landing scenarios, with an average final position error of 9.0 cm and 11 cm for stationary and mobile platforms, respectively. Notably, NMPC Lander outperforms the B-spline combined with the A* planning method by nearly threefold in terms of position tracking, underscoring its superior robustness and practical effectiveness.

Keywords: Human-Computer Interaction, Human-centered Learning
Abstract: ，基于元学习和对比 基于学习的方法取得了可喜的成果 在解决建议中的冷启动问题时 通过全局建模用户偏好来构建系统。然而，由于 冷启动之间稀疏的历史交互 物品和用户，很难准确捕捉 用户和项目之间的层次结构关系。自 为了缓解上述问题，我们提出了一种双空间映射 冷启动推荐的对比元学习 （DMCM），同时映射用户和项目信息 进入庞加莱空间和欧几里得空间到更多 有效捕捉 用户-项目交互信息。通过双空间 映射对比学习，表征能力 用户-项目关联的覆盖率得到增强，并且 展开用户交互信息。随后，一个 多层次元学习路径动态调整策略 用于进一步提高模型能力 个性化并适应新用户和新项目。 在三个公共数据集上进行的实验表明 所提

Keywords: Human-Computer Interaction, Human-Machine Interaction
Abstract: Wearable electromyography (EMG)-based control interfaces are widely used by various users including both individuals with and without physical disabilities. Since EMG-based control interfaces are unfamiliar to most users and require individually defined input detection thresholds, differences in the sense of agency among users are to be expected. This study aimed to investigate individual trends in the sense of agency when operating an unfamiliar EMG-based control interface. To this end, we utilized a task that combines the intentional binding (IB) measurement paradigm with a perceptual-motor learning task to measure task performance and implicit sense of agency during EMG operation. To classify individual trends in the sense of agency, we conducted clustering based on variations in raw IB data. We found that cluster with large variations in IB had a low implicit sense of agency, reflecting the difficulty of the interface operation in detecting muscle movements as input. In addition, an analysis of the cluster with low variability in IB suggests that participants who took longer to become familiar with the input operation exhibited higher implicit agency, while those who found the task easy exhibited lower implicit agency. These findings suggest that IB may serve as a useful measure for capturing differences in operability not reflected in task performance or explicit reports of sense of agency.

Keywords: Human-Computer Interaction, User Interface Design, Human-Collaborative Robotics
Abstract: In this paper, we present a systematic method of design for human-swarm interaction interfaces, combining theoretical insights with empirical evaluation. We first derived ten design principles from existing literature, applying them to key information dimensions identified through goal-directed task analysis and developed a tablet-based interface for a target search task. We then conducted a user study with 31 participants where humans were required to guide a robotic swarm to a target in the presence of three types of hazards that pose a risk to the robots: Distributed, Moving, and Spreading. Performance was measured based on the proximity of the robots to the target and the number of deactivated robots at the end of the task. Results indicate that at least one robot was brought closer to the target in 98% of tasks, demonstrating the interface’s success in fulfilling the primary objective of the task. Additionally, in nearly 67% of tasks, more than 50% of the robots reached the target. Moreover, particularly better performance was noted in moving hazards. Additionally, the interface appeared to help minimise robot deactivation, as evidenced by nearly 94% of tasks where participants managed to keep more than 50% of the robots active, ensuring that most of the swarm remained operational. However, its effectiveness varied across hazards, with robot deactivation being lowest in distributed hazard scenarios, suggesting that the interface provided the most support in these conditions.

Keywords: Human-Machine Interaction, Human-Computer Interaction, Human-Machine Cooperation and Systems
Abstract: This study conducts a comparative evaluation of the YOLO and DETR models as an initial step toward the development of a cost-effective and accurate object detection system for images of character-themed products, such as stuffed animals, on social media platforms. The evaluation focuses on the detection methods and accuracy of each model, analyzing and identifying the strengths and characteristics of each model.

Keywords: Human-Computer Interaction, Human-Machine Interface, Human-Machine Interaction
Abstract: Providing effective indoor spatial guidance through natural language remains a key challenge in human-computer interaction. This paper presents Orlock, a virtual agent designed to support wayfinding and spatial orientation within a university building. To ensure a consistent baseline, building maps were shown to participants unfamiliar with the space. Following their interaction with Orlock, participants completed an 18-item questionnaire covering five key dimensions: clarity, usefulness, ease of interaction, overall satisfaction, and design preferences. Results from the 7-point Likert-scale responses indicated generally high satisfaction, with a clear preference for multimodal feedback combining visual and verbal instructions. These findings highlight the importance of multimodal interaction in orientation systems and offer practical design insights for future development of intelligent spatial guidance agents.

Keywords: Human-Computer Interaction, Multi-User Interaction, Assistive Technology
Abstract: Over the past decade, the demand for communication devices has increased among individuals with mobility and speech impairments. Eye-gaze tracking has emerged as a promising solution for hands-free communication; however, traditional appearance-based interfaces often face challenges such as accuracy issues, involuntary eye movements, and difficulties with extensive command sets. This work presents a multimodal appearance-based gaze-controlled virtual keyboard that utilises deep learning in conjunction with standard camera hardware, incorporating both synchronous and asynchronous modes for command selection. The virtual keyboard application supports menu-based selection with nine commands, enabling users to spell and type up to 56 English characters—including uppercase and lowercase letters, punctuation, and a delete function for corrections. The proposed system was evaluated with twenty able-bodied participants who completed specially designed typing tasks using three input modalities: (i) a mouse, (ii) an eye-tracker, and (iii) an unmodified webcam. Typing performance was measured in terms of speed and information transfer rate (ITR) at both command and letter levels. Average typing speeds were 18.3±5.31 letters/min (mouse), 12.60±2.99 letters/min (eye-tracker, synchronous), 10.94±1.89 letters/min (webcam, synchronous), 11.15 ± 2.90 letters/min (eye-tracker, asynchronous), and 7.86 ± 1.69 letters/min (webcam, asynchronous). ITRs were approximately 80.29 ± 15.72 bits/min (command level) and 63.56 ± 11 bits/min (letter level) with webcam in synchronous mode. The system demonstrated good usability and low workload with webcam input, highlighting its user-centred design and promise as an accessible communication tool in low-resource settings.

Keywords: System Modeling and Control
Abstract: In this paper, we delve into the examination of periodic dynamics, specifically with periods one and two, exhibited by closed-loop DC-DC buck converters operating in continuous conduction mode (CCM). Our approach employs a novel linear theoretical method introduced by the authors. The investigation utilizes state sequence flow charts derived from this innovative linear method to analyze the periodic dynamics inherent in closed-loop CCM DC-DC buck converters. Our research not only establishes theoretical foundations for the general exact closed-form solutions to system states and duty cycles but also validates them through simulation, enhancing the confirmation of our findings. The simulation verification assesses various aspects, including system states, duty cycles, average output voltages, and output voltage ripples, considering variations in input voltages. In essence, our work presents a straightforward and precise methodology to unravel the intricate nonlinear characteristics embedded in closed-loop CCM DC-DC buck converters.

Keywords: System Modeling and Control, Communications, Control of Uncertain Systems
Abstract: Reliable low-latency media streaming is increasingly critical for delivering seamless interactive and immersive services. To meet this need, the IETF and 3GPP have introduced the Low Latency, Low Loss, Scalable Throughput (L4S) architecture, which mitigates queuing delays in IP traffic and supports latency-sensitive applications. This paper focuses on real-time video streaming and proposes an enhanced framework, Enhanced L4S (EL4S), which integrates a link load factor feedback mechanism to more accurately reflect real-time network conditions.We design and implement a cross-layer end-side transmission control algorithm based on EL4S to improve real-time video performance over 5G networks. In this design, EL4S encodes link load information into packets at the 5G core and uses ACK-based feedback to inform the sender about current network states. At the sender, a dynamic bitrate adaptation algorithm adjusts the transmission rate in response to the reported link load factor, balancing throughput and delay while avoiding network congestion. This adaptive mechanism enables precise, frame-level control over video encoding rates and transmission behavior.Extensive experimental evaluations demonstrate that EL4S achieves high link utilization and low latency, significantly outperforming existing baseline algorithms. Overall, EL4S provides an efficient and deployable solution for achieving low-latency, high-quality real-time streaming in dynamic 5G networks.

Keywords: System Modeling and Control, Consumer and Industrial Applications, Robotic Systems
Abstract: In computer vision, most data are captured in 2D formats, limiting spatial understanding in real-world applications. This presents a challenge for fields such as architecture, construction, and robotics, where interpreting spatial relationships from minimal visual input is increasingly essential. This survey reviews recent advancements in extracting 3D features from 2D imagery, a critical task in these domains, where spatial accuracy and object orientation directly impact performance. We focus on three core areas: (1) disposition estimation, determining object pose; (2) joint modeling, constructing skeletal representations; and (3) scene reconstruction, generating spatially accurate environments. Each category is evaluated based on input modalities, performance metrics, and code availability. By providing a unified overview of these techniques, this paper highlights their practical value in enabling 3D reasoning from conventional 2D data.

Keywords: System Modeling and Control, Cyber-physical systems, Intelligent Power Grid
Abstract: Virtual Power Plants (VPPs) have been increasingly being used to achieve carbon neutrality in energy systems and to improve resilience. This study handled a control scheme for small-scale community-driven VPPs, which has attracted much attention in recent years. In such community-driven VPPs, operations are not limited to maximizing economic value, but are also focused on community preferences. In this study we proposed Model Predictive Allocation Control (MPAC), which enables VPPs to operate in a way that appropriately reflects community preferences. The MPAC formulates the energy resource allocation control by Model Predictive Control (MPC) and modifies the VPP operation by tuning the weight parameters of the MPC. Furthermore, the weight parameters can be tuned by a preference learning-based optimization algorithm to easily reflect the community's decisions. We also compensate for the operational stability of the VPPs by using frequency stabilizing control in combination. The effectiveness of the proposed method was verified by experiments using numerical simulations.

Keywords: System Modeling and Control, Decision Support Systems, Intelligent Transportation Systems
Abstract: The study focuses on the multi-UAV assignment network system problem considering route planning. It firstly establishes geographic environment model and designs the cost function and the flight constraints of the UAV in the transport task process. Combined with the particle swarm optimization algorithm, it uses simulation software to solve the assignment network system problem and constructs the transportation assignment network with complete route information on the basis of the optimal flight route generated by the algorithm. According to the relevant actual scenes, the transportation network simulation experiment and the multi-point distribution simulation experiment are carried out respectively, so as to obtain the optimal assignment planning scheme that accomplishes the cost objective of the multi-UAV under the flow restriction and transportation point requirements.

Keywords: System Modeling and Control, Decision Support Systems, Intelligent Transportation Systems
Abstract: Maritime networks constitute a critical pillar of the global economic system and reflect intricate trade relationships between nations. Meanwhile, the application of complex network theory has increasingly highlighted the intrinsic link between node influence and dependencies among non-adjacent nodes. However, a central challenge lies in uncovering latent dependencies—mediated by intermediaries—between non-directly connected nations, which is crucial for the precise modeling of maritime networks and the analysis of indirect influence. In this study, we construct and optimize a national-level multi-order influence network by mining multi-order dependency rules through advanced network construction methods. Empirical analysis of real-world ship navigation trajectory data (August 2018 and August 2023) reveals increasingly interconnected global dependencies. The proposed method successfully quantifies multi-order influence, identifies strategic hubs, tracks the evolution of influence, and offers actionable insights into geopolitical shifts and strategies for trade resilience.

Keywords: Human-Centered Transportation, Design Methods, Human-Machine Interaction
Abstract: Vehicle driving simulators have been widely used in fields including road design, automotive development, and driver training. As a core component of the simulators, motion cueing algorithms (MCAs) aim to reproduce realistic vehicle motion sensations while respecting workspace limitations for a high-fidelity driving experience. Recent advancements have identified model predictive control (MPC) as a promising approach for MCA. However, conventional MPC-based MCA faces computational bottlenecks that limit real-time performance. Although recent studies have employed swarm intelligence optimization algorithms like genetic algorithm and grey wolf optimizer to enhance the performance of MPC-based MCA through horizon adjustments, there was still an overlook of improvement path from the computational cost perspective. Therefore, this paper proposes a costs-optimization framework for MPC-based MCA (COMPC-based MCA), which integrates two key components: an Operator Splitting Quadratic Program (OSQP) solver and a Snow Ablation Optimizer (SAO) that considers computational costs and sensation errors for parameters selection. Experiment results validated the advantages of the proposed framework. COMPC-based MCA can achieve fast quadratic programming (QP) solving and select appropriate parameters for performance improvement while considering computational costs.

Keywords: Human-Centered Transportation, Human Factors, Human-Collaborative Robotics
Abstract: In this paper, first of all, the group walking behavior by four pedestrians are observed. In the observation, not only the motion data but also the decision making of each pedestrian are collected by using special device. Then, three behavioral indicators: deceleration, detour amount, and decision entropy, are defined and calculated. It has been found that these three indicators successfully quantify the ’smoothness’of the group walking behavior. Finally, the principal component analysis(PCA) is applied to the three dimensional indicator data. As the result, the meaning of three principal components are clearly explained. The discussion based on the PCA will be a basis for the further analysis and classification of the group walking behavior.

Keywords: Human-Centered Transportation, Human Factors, Human-Computer Interaction
Abstract: Early identification of driving risks is crucial for driving safety. In this study, a deep learning method based on Long Short-term Memory (LSTM) with attention mechanism and XGBoost to recognize risk levels during driving simulation for early warnings, has been proposed. Here, we collected multimodal data -including driving data, eyes status, and basic physiological data-from driving simulators, eye trackers, and smartwatches for comprehensive analysis. By combining driving behavior and physiological features in each time window, the ability of LSTM networks was applied to analyze temporal features for prediction of the risk levels (Level 0, Level 1, and Level 2) in the next 20 seconds. Meanwhile, an attention mechanism was introduced to understand the importance of multiple features as well. Then the output of LSTM-attention was further refined using XGBoost, which demonstrated excellent performance in classification tasks. The integrated model showed that the overall accuracy achieved is above 89.97%, with an F1-score of 89.15%.

Keywords: Human-Centered Transportation, Human-Computer Interaction, Information Systems for Design
Abstract: As urban centers face growing challenges from population growth, traffic congestion, and environmental degradation, sustainable mobility solutions have become essential. This paper presents a performance evaluation of a public bicycle-sharing system using Stochastic Petri Nets (SPNs) to model and simulate user behavior. Grounded in recent research on urban mobility and shared transportation, the proposed stochastic framework captures key operational parameters, including bicycles in transit, bicycles at stations, waiting probability, system utilization, and the availability of parking docks. A case study evaluates three high-demand stations under varying demand scenarios (baseline, − 25%, and +25%). Results indicate that, while the system operates adequately under current demand, it is prone to saturation under increased usage and exhibits inefficiencies when demand decreases. The proposed model provides valuable insights to support operational improvements, infrastructure planning, and public policy formulation for integrated and sustainable urban mobility.

Keywords: Evolutionary Computation, Computational Intelligence
Abstract: In the community of evolutionary multi-objective optimization (EMO), large-scale multi-objective optimization problems (LSMOPs) with many decision variables have attracted much attention. The main difficulty of LSMOPs lies in their high-dimensional decision space, which slows down the convergence of EMO algorithms towards the Pareto front. To address this issue, many novel variation operators have been proposed to improve the efficiency of EMO algorithms. However, for both conventional EMO algorithms (e.g., NSGA-II) and recently proposed EMO algorithms (e.g., LERD), the polynomial mutation with the mutation probability 1/n, where n is the number of decision variables, is always used. For LSMOPs with a large number of decision variables, the mutation probability 1/n looks too small (e.g., 1/1000). In this paper, we examine different mutation probabilities and find that many existing EMO algorithms with a larger mutation probability (e.g., 10/n) are significantly better than the standard setting (i.e., 1/n) in handling LSMOPs.

Keywords: Evolutionary Computation, Computational Intelligence, Heuristic Algorithms
Abstract: Momentum is a technique that adds the momentum moves from the earlier iterations into the current update to accelerate convergence. While the momentum technique has been widely used in single-objective optimization, its application in evolutionary multi-objective optimization (EMO) has not gained much attention. Since EMO algorithms are population-based algorithms, how to choose solutions for applying momentum becomes an important issue. Inspired by Polyak's momentum method and Nesterov's momentum method in single-objective optimization, we propose four different momentum methods for EMO. Our findings demonstrate that the performance of EMOAs with momentum is strongly affected by the choice of solutions to which momentum moves are applied.

Keywords: Evolutionary Computation, Computational Intelligence, Swarm Intelligence
Abstract: Crowdsourcing has become a powerful tool for data collection and problem-solving, harnessing collective intelligence to address complex tasks. Building knowledge bases through such intelligence is gaining attention to enhance crossdomain model performance. As an essential technique for dimensionality reduction, the task of feature selection (FS) also emerges in the context of crowdsourcing, resulting in crowdsourcing feature selection (CFS). Due to the diversity of workers and the varying quality of data in a crowdsourcing environment, CFS faces challenges such as heterogeneous data reliability and dynamic participation. To tackle these issues, we present Crowd-Fed Evolutionary Transfer Optimization (CFETO), an efficient and crowdsourcing-enhanced framework for feature selection. CFETO comprises a central server and multiple distributed workers: workers perform local data collection and optimization, extracting knowledge from their datasets, while the server integrates this knowledge to build a knowledge base. An evolutionary transfer learning strategy is further employed to harness this knowledge, thereby improving both convergence speed and selection robustness. Experiments on 20 real-world datasets show that CFETO outperforms traditional centralized FS methods approaches, underlining its potential for broader application in complex, distributed crowdsourcing scenarios.

Keywords: Evolutionary Computation, Computational Intelligence, Swarm Intelligence
Abstract: Surrogate model-assisted evolutionary algorithm (SAEA) has become a generalized method for expensive optimization problems (EOPs) with high-cost evaluations. However, most existing SAEAs retrain surrogate models frequently during evolutionary iterations. Besides, the quality of sample data directly affects model accuracy, leading to difficulties in finding the optimal solution. To address these challenges, this paper introduces the tree-based broad learning system (TBLS) as the surrogate model into SAEA framework, and a TBLS-assisted optimizer with incremental learning (TBLSOIF) is proposed. First, an incrementable TBLS is constructed for predicting the quality of the solution, and the model is incrementally updated by expanding the layer nodes when the samples increase. And then effectively avoids the high computational costs of retraining models from scratch and significantly improves the update efficiency of surrogate models. Second, an adaptive sample augmentations (ASA) strategy is designed to generate more samples for updating the TBLS using a variational autoencoder (VAE). Experiments on 15 problems in the CEC2017 benchmark functions show that the proposed TBLSO-IF algorithm is more effective and competitive than the other five state-of-the-art SAEA methods.

Keywords: Evolutionary Computation, Computational Intelligence, Swarm Intelligence
Abstract: Evolutionary multi-objective optimization (EMO) algorithms generate diverse Pareto optimal solutions to comprehensively approximate the Pareto front. However, an excessive number of diverse solutions pose significant challenges for decision makers who must select a single or few solutions for implementation. To address this decision-making burden, we propose a novel approach that promotes consistency among decision variables while accepting controlled performance trade-offs. Specifically, we develop VCM2M, an EMO algorithm that integrates a variable consistency strategy into the MOEA/D framework. It decomposes a multi-objective optimization problem (MOP) into a number of subproblems. In each subproblem, we incorporate the logarithmic sum regularization term of the decision variables into objective functions, achieving consistency during both the optimization and selection phases. An epsilon-insensitive loss function is utilized to normalize the loss of population diversity in the objective space. Experimental evaluation on specially constructed test problems demonstrates that VCM2M successfully generates solution sets organized into distinct clusters, where solutions within each cluster exhibit significantly higher variable consistency compared to those produced by MOEA/D-M2M, AR-MOEA, and NSGA-III.

Keywords: Evolutionary Computation, Heuristic Algorithms, Optimization and Self-Organization Approaches
Abstract: The hypervolume (HV) indicator is widely used in multi-objective evolutionary algorithms (MOEAs) due to its Pareto compliance property. This property makes it very effective for assessing the quality of approximation sets from different MOEAs and for ranking solutions among a population of solutions using the individual hypervolume contribution (HVC). However, the computational cost of computing the HV increases exponentially with the number of objectives. Furthermore, this increase in computational cost is worse when adopting the HVC as a density estimator, making it prohibitive in many-objective optimization problems (MaOPs). In this work, we propose a novel approach to create HVC approximation functions using Grammatical Evolution (GE). We describe the grammar and fitness functions designed to identify the worst-contributing individual given a population of non-dominated solutions. Then, we use a GE implementation with training data generated from the DTLZ and WFG benchmark problems. The resulting approximation functions, tailored for dimensionalities ranging from 2 to 10, are evaluated against two state-of-the-art methods: HVC-Net and the R2-based HVC approximation. Experimental results on validation data also derived from benchmark problems show that our GE-generated functions consistently outperform both alternative approaches regarding worst-contributing individual identification for dimensions greater than two while maintaining competitive execution times. These results indicate that GE is a viable and effective tool for generating high-quality HVC approximations, particularly suitable for solving MaOPs.

Keywords: Cyber-physical systems
Abstract: Sound Event Detection (SED) aims to identify and locate specific sound events within audio streams. Despite recent advancements, current methodologies exhibit two fundamental limitations: (1) Insufficient modeling of discriminative temporal-spectral characteristics induces compromised differentiation for acoustically similar events. (2) Failing to address both inter-class and intra-class imbalance problems induces biased classifications toward dominant classes and inactive frames. To tackle these challenges, we propose Acoustic Clustering and Imbalance Learning-based Sound Event Detection (ACIL-SED), which consists of Cluster-Specialized Convolutional Recurrent Neural Network (CS-CRNN) and Dual-Objective Adaptive Balance Loss (DOABLoss). The CSCRNN employs acoustic clustering-guided specialized submodels, where each cluster’s sub-model focuses on discriminative feature learning in specific temporal-spectral characteristics, thereby resolving the compromised feature representation inherent in a shared single-model architecture. The DOABLoss adaptively combines mean squared error (MSE) and weighted binary cross-entropy (WBCE) to simultaneously address the issues of inter-class duration imbalance and intra-class activeinactive frame imbalance. Experimental results show that ACIL-SED outperforms state-of-the-art methods in complex acoustic environments.

Keywords: Cyber-physical systems, Control of Uncertain Systems, Adaptive Systems
Abstract: This paper studies a data-driven event-triggered sliding-mode control problem for wind turbines with prescribed performance and quantified information to maximize power generation efficiency. Initially, a partial form dynamic linearization model is established for the controlled wind turbine system. A logarithmic quantizer is considered to quantize data before it is transmitted. Then, a data-driven event-triggered sliding-mode control approach is established, where a smooth function is introduced to limit the control error converges to a prescribed range, and an event-triggered scheme is designed to reduce the communication frequencies of the controlled plant. Finally, the convergence of the proposed method is rigorously proven, and the effectiveness of the proposed method is demonstrated through simulation studies.

Keywords: Cyber-physical systems, Digital Twin, System Modeling and Control
Abstract: In the synchronization of cyber-physical systems (CPSs), modeling the nonlinear dynamics of physical plants is a challenging task. To address this challenge, we propose a novel H∞ controller design method that leverages a data-driven approach to robustly synchronize CPSs and ensure their stability. In the proposed approach, the input-output relationship of the physical system is learned using long short-term memory (LSTM) networks to approximate the unknown dynamics of CPSs. Furthermore, we exploit an effective control scheme for trained LSTM networks to effectively handle the nonlinearity of activation functions. To ensure stability and performance in the convergence of synchronization error, a controller design criterion is derived for the trained LSTM network in terms of linear matrix inequalities, and the controller gain is computed using convex optimization techniques. In addition, we present an anomaly detection algorithm using the proposed method, which can synchronize CPSs and detect abnormal signals without requiring any prior physical model information. Consequently, the stability of the synchronization control system can be ensured, enabling its application to anomaly detection. Finally, the effectiveness of the proposed method is validated through an experiment on a motor control system even in abnormal operating conditions.

Keywords: Cyber-physical systems, Robotic Systems, Smart Buildings, Smart Cities and Infrastructures
Abstract: Providing Artificial Intelligence as a Service (AIaaS) for the next generation of Industry, known as Industry 5.0, is committed to providing an accurate indoor localization system. Providing Indoor localization, particularly in dynamic industrial environments where operational efficiency relies on precise location information, is a challenging task. This paper presents a Collaborative Indoor Positioning System (CIPS) facilitated by a location-aware autonomous mobile robot equipped with Bluetooth Low Energy (BLE) beacons to enhance indoor localization accuracy through dynamic path loss parameter estimation. The robot traverses the facility while periodically broadcasting its ID and real-time location to a central server. By correlating the robot’s ground-truth positions with Received Signal Strength Indicator (RSSI) measurements from fixed BLE receivers, the server continuously updates path loss parameters using a log-normal shadowing model and linear regression. Evaluation of the proposed system is conducted using a real testbed with a deployed application system in an office environment. Experimental results demonstrate a significant improvement in distance estimation accuracy, ranging between 19% and 32%, and a 20% reduction in location estimation error compared to conventional methods, highlighting the potential of the proposed CIPS approach for enhancing indoor localization in industrial settings.

Keywords: Cyber-physical systems, Discrete Event Systems
Abstract: In this paper, the problem of cyber-attacks detection in timed probabilistic discrete event systems via artificial neural networks is investigated. We extend the problem of state estimation for timed probabilistic discrete event systems using artificial neural networks, to address attack detection. So that, a detection strategy is implemented to determine whether the system is operating in normal mode or under attack, and to identify the potential type of attack. Two primary cases are examined: (i) attack detection over observations. In this case, the attack detector recognizes whether the system is in a normal mode or an attack mode after each new observation, (ii) attack detection over time. Here, the attack detector evaluates the system’s status at each clock time increment.

Keywords: AI and Applications, Application of Artificial Intelligence
Abstract: Level 3 automated driving systems allows drivers to engage in secondary tasks while diminishing their perception of risk. In the event of an emergency necessitating driver intervention, the system will alert the driver with a limited window for reaction and imposing a substantial cognitive burden. To address this challenge, this study employs a Large Language Model (LLM) to assist drivers in maintaining an appropriate attention on road conditions through a “humanized” persuasive advice. Our tool leverages the road conditions encountered by Level 3 systems as triggers, proactively steering driver behavior via both visual and auditory routes. Empirical study indicates that our tool is effective in sustaining driver attention with reduced cognitive load and coordinating secondary tasks with takeover behavior. Our work provides insights into the potential of using LLMs to support drivers during multi-task automated driving.

Keywords: Human Factors, Augmented Cognition
Abstract: Successful decision-making in uncertain contexts depends on balancing comfortable, tried-and-tested options with curious exploration. Yet, high cognitive load tends to disrupt this equilibrium, lowering risk-taking and exploratory responses. Although neuromodulation through transcranial direct current stimulation (tDCS) has been shown to boost cognitive flexibility, its combination with cognitive load in influencing curiosity is not well understood. This study investigates how anodal tDCS applied to the dorsolateral prefrontal cortex affects risk-taking and exploration under different cognitive loads. Thirty participants were assigned to one of three conditions: (1) tDCS with trivia- based cognitive load, (2) trivia without tDCS, and (3) no-load, no-tDCS stimulation (control). Participants performed a 50-trial decision-making task with three options per trial: a certain safe option, a risky high-reward option, and an exploratory option that provided probabilistic information.

Measures of behavior were risk-taking, exploratory behavior, and cognitive flexibility in terms of strategy switching. Outcome showed that the tDCS with trivia condition had significantly more risk-taking and exploration compared to both control groups. Importantly, even without cognitive load, there was a facilitation of exploration. Cognitive flexibility was most pronounced for the tDCS group and implies that neuromodulation can support adaptation behavior under load. These observations prove that a decrease in cognitive load as well as neuromodulation separately and interactively improves decision making, curiosity, and risk acceptance. This implies real- world value in high-demand areas like healthcare, education, and aviation where preserving exploration when under pressure is paramount.

Keywords: Information Assurance and Intelligence, Application of Artificial Intelligence, Deep Learning
Abstract: Aerial search and rescue (SAR) rotorcrafts currently need multiple specialist human operators, increasing cost and the risk of downtime due to crew unavailability and mental stress. Autonomy can aid fewer operators performing multiple tasks, but the human operator must maintain situation awareness (SA) of crucial autonomous decisions. A key challenge is the cognitive stress on a multi-tasking human-in-the-loop (HITL) due to the AI agent making decisions without human understanding. Explainable AI (XAI) has often been proposed as a way to explain autonomy decisions, but current XAI solutions doesn't adapt to real-time human factors in high stress and high stakes situations. Here, we allow an AI agent to perform autonomous rotorcraft navigation, whilst the HITL operator has to perform two simultaneous tasks: (i) search for a target on the ground by toggling an onboard camera, and (ii) maintain SA of the autonomous navigation task through our novel XAI interface. Our novel XAI approach leverages on dimensionality reduction techniques to visualize the reinforcement learning (RL) navigation's internal states, highlighting patterns in its decision-making process through intuitive interactive clustering on saliency maps. To ensure convergence on performance, we design a two-way interface that allows the human to interpret AI decisions and then give feedback via a Large Language Model to modify the autonomous navigation. Testing demonstrates increased task performance (+43%), while experiencing substantial human reductions in physical demand (-53%), time pressure (-30%), effort (-23%), and frustration (-26%), but at the cost of slightly increased mental demand (+12%).

Keywords: Computational Intelligence, Machine Vision, Computational Intelligence in Information
Abstract: Landslide early warning systems (EWS) are critical to disaster preparedness but are frequently limited by uncertainty of prediction and variability in user trust. This research presents a new dual-modality method combining virtual reality (VR) simulation and electroencephalography (EEG) to evaluate behavioral and neurophysiological reactions to probabilistic landslide warnings. Eighty drivers experienced a VR driving situation with different warning accuracy (70% vs. 95%) and lighting conditions (day vs. night), collecting behavioral measures (e.g., collisions, speed, trajectory deviance) and EEG-based cognitive measures (e.g., alpha/theta, alpha/gamma, beta/gamma ratios) as dependent measures. Results revealed that decreased warning accuracy caused elevated collision rates, route deviances, and beta/gamma EEG activity, representing higher cognitive stress. Higher ratios of alpha/theta and alpha/gamma were related to performance in driving and were more prominent under higher accuracy and daylight. These results stress the promise that neuroadaptive VR systems hold to improve disaster training by dynamically calibrating feedback according to the cognitive states of users, therefore providing useful insights into the intelligent, human-orientated EWS technology design within the fields of system, man, and cybernetics.

Keywords: AIoT, Neural Networks and their Applications, Application of Artificial Intelligence
Abstract: In this study, a recurrent spiking neural network (RSNN) has been implemented on an accelerator deployed on a reconfigurable circuit (FPGA) to predict the Condenser Fouling Factor (CFF) in a chiller as a predictive maintenance (PdM) measure. This system operates through a low-power device utilising edge computing, geared towards AIoT/EdgeAI applications. It can accurately identify CFF levels that exceed 25% based on several pressure and temperature sensor data distributed in the chiller, achieving an accuracy greater than 90% with an architecture of 256 parallel and recurrent neurons in the hidden layer. The model was trained using a proprietary dataset that recorded sensor states during controlled experiments where the condenser was manually obstructed at varying coverage percentages. The primary advantage of employing RSNN techniques lies in their dual capability: first, they are designed to detect temporal signal patterns, and second, their trainability allows for adaptation to various applications across different contexts. The training for the particular accelerator used in this work is done on the FPGA, not requiring power-hungry machines.

Keywords: Cloud, IoT, and Robotics Integration, Knowledge Acquisition, Intelligent Internet Systems
Abstract: The transformation of the energy sector into a cyber-physical system of systems, driven by the integration of renewable energy sources and digitalization, demands new engineering and validation methodologies. Traditional monolithic approaches lack the flexibility, interoperability, and collaborative capabilities required for modern power system automation applications. This work introduces the PowerTeams approach, a cloud-native, service-oriented platform designed to support collaborative engineering across the entire lifecycle of power system automation applications. The platform integrates both collaboration and engineering services, supports dynamic service registration, and employs a microservice architecture with container-based deployment to ensure scalability and flexibility. A validation scenario involving the development and deployment of a power plant controller for a photovoltaic system demonstrates the platform’s capabilities. Initial evaluations with external users confirm the feasibility and benefits of the approach, highlighting its potential for real-world power and energy system projects.

Keywords: Cyber-physical systems, Consumer and Industrial Applications, Decision Support Systems
Abstract: The increasing complexity of multi-plant manufacturing systems demands architectural models able to support coordinated, scalable, and secure industrial operations. This paper introduces a five-layer architecture for distributed smart manufacturing environments to enhance interoperability, standardization, and cyber-physical integration across multiple facilities. The architecture leverages Digital Twin technology to synchronize physical operations with digital models, enabling coordinated control, monitoring, and decision support. Its layered structure supports modular deployment across plant-level and enterprise systems, while ensuring compatibility with heterogeneous technologies and legacy infrastructures. Specific emphasis is placed on cross-site data aggregation and external information sharing in compliance with data space principles. A prototype implementation validates the applicability of the architecture and its ability to support scalable and secure industrial operations, laying the groundwork for broader adoption in interconnected manufacturing ecosystems.

Keywords: Digital Twin, Cyber-physical systems, Consumer and Industrial Applications
Abstract: The increasing complexity of Cyber-Physical Systems (CPS), particularly in the industrial domain, has amplified the challenges associated with the effective integration of Artificial Intelligence (AI) and Machine Learning (ML) techniques. Fragmentation across IoT and IIoT technologies, manifested through diverse communication protocols, data formats and device capabilities, creates a substantial gap between low-level physical layers and high-level intelligent functionalities. Recently, Digital Twin (DT) technology has emerged as a promising solution, offering structured, interoperable and semantically rich digital representations of physical assets. Current approaches are often siloed and tightly coupled, limiting scalability and reuse of AI functionalities. This work proposes a modular and interoperable solution that enables seamless AI pipeline integration into CPS by minimizing configuration and decoupling the roles of DTs and AI components. We introduce the concept of Zero Configuration (ZeroConf) AI pipelines, where DTs orchestrate data management and intelligent augmentation. The approach is demonstrated in a MicroFactory scenario, showing support for concurrent ML models and dynamic data processing, effectively accelerating the deployment of intelligent services in complex industrial settings.

Keywords: Digital Twin, Cyber-physical systems, Robotic Systems
Abstract: Support point queries are a critical part of many collision detection pipelines, including those for robotics and real-time graphical applications. This paper proposes several memory layout optimizations to speed up support point queries on convex hulls. These methods are implemented and tested on a variety of different hardware models, with a decrease in processing time of up to five times compared to current approaches. The results in this paper can be integrated with existing physics libraries with minimal effort.

Keywords: Digital Twin, Discrete Event Systems, System Modeling and Control
Abstract: Process Mining (PM) has been proven valuable for extracting process flows from data, also in the form of stochastic Petri net (SPN) models of systems. SPNs are widely recognized for their ability to model complex, stochastic systems and are extensively used in combination with PM. While SPNs provide an intuitive and straightforward way to model complex systems, representing changes across multiple dimensions, such as energy and waste, remains challenging in their standard frameworks. In this paper, we introduce an extension of stochastic Petri nets, termed Multidimensional SPNs (MDSPNs), by extending the SPN framework to capture dynamics along different dimensions. MDSPNs facilitate a comprehensive modeling of systems’ behaviors from multiple perspectives, which can correspond to the diverse objectives of systems. To facilitate design and simulation of MDSPNs, we designed and developed MDPySPN, a Python library, which we also introduce in this paper. MDPySPN enables the simulation of MDSPNs by supporting alterations of multiple values at system events. With MDPySPN, we aim to provide researchers, engineers, and simulation professionals with a practical and extensible toolkit to model, simulate, and analyze MDSPNs, thereby supporting multi-objective optimization of stochastic processes in systems. Through a case study, we demonstrate the capabilities of modeling and simulation of MDSPNs using MDPySPN.

Keywords: Digital Twin, Manufacturing Automation and Systems, Fault Monitoring and Diagnosis
Abstract: In subtractive manufacturing, predicting the life- span of tools offers significant advantages. However, data-driven methods for such predictions often fail to incorporate domain knowledge effectively. This paper presents a hybrid approach that combines machine learning with domain expertise. Firstly, we introduce a novel method to augment sparsely labeled datasets. Secondly, we propose a new loss function that inte- grates domain knowledge into a machine learning model. This approach enhances the accuracy and reliability of tool wear predictions, ultimately improving the efficiency of subtractive manufacturing processes. We evaluate the proposed methods on the NUAA Ideahouse Dataset. We achieve a R2 score of up to 0.99 on unseen data.

Keywords: Digital Twin, Technology Assessment
Abstract: The digital twin of humans is a relatively new concept. While many diverse definitions, architectures, and applications exist, a clear picture is missing on what, in fact, makes a human digital twin. Within this context, researchers and industrial use-case owners alike are unaware about the market potential of the - at the moment - rather theoretical construct. In this work, we draw a holistic vision of the human digital twin, and derive the specification of this holistic human digital twin in form of requirements, stakeholders, and users. For each group of users, we define exemplary applications that fall into the six levels of functionality: store, analyze, personalize, predict, control, and optimize. The functionality levels facilitate an abstraction of abilities of the human digital twin. From the manifold applications, we discuss three in detail to showcase the feasibility of the abstraction levels and the analysis of stakeholders and users. Based on the deep discussion, we derive a comprehensive list of requirements on the holistic human digital twin. These considerations shall be used as a guideline for research and industries for the implementation of human digital twins, particularly in context of reusability in multiple target applications.

Keywords: Neural Networks and their Applications, Computational Intelligence, Image Processing and Pattern Recognition
Abstract: Federated Learning (FL) is a promising machine learning paradigm that enables participating devices to train privacy-preserved and collaborative models. FL has proven its benefits for robotic manipulation tasks. However, grasping tasks lack exploration in such settings where robots train a global model without moving data and ensuring data privacy. The main challenge is that each robot learns from data that is nonindependent and identically distributed (non-IID) and low quantity. This exhibits performance degradation, particularly in robotic grasping. Thus, in this work, we propose MTF-Grasp, a multi-tier FL approach for robotic grasping, acknowledging the unique challenges posed by the non-IID data distribution across robots, including quantitative skewness. MTF-Grasp harnesses data quality and quantity across robots to select a set of ``top-level'' robots with better data distribution and higher sample count. It then utilizes top-level robots to train initial seed models and distribute them to the remaining ``low-level'' robots, reducing the risk of model performance degradation in low-level robots. Our approach outperforms the conventional FL setup by up to 8% on the quantity-skewed Cornell and Jacquard grasping datasets.

Keywords: Machine Learning, Swarm Intelligence, Intelligent Internet Systems
Abstract: Modern applications increasingly rely on knowledge and collective wisdom. While the internet offers abundant sources, it is untrustworthy, incomplete, and essentially a vast, heterogeneous database lacking current information. With innovation accelerating exponentially, harnessing collective intelligence is essential to drive progress. We present emph{Consensus Without Authority}, a general, ledger-backed meta-protocol framework that transforms any group of self-interested actors into a trusted collective-intelligence engine. Each actor i produces a local insight ( LI_i = f_1(mathcal{Q}, theta_i) ), evaluates its peers ( U_i = f_2({LI_j}) ), and enters a two-phase commit-reveal cycle. Harmonizers aggregate votes ( V = f_3({U_i}) ), synthesize candidate knowledge ( CK_j = f_4(V, {LI_i}) ), and a smart contract finalizes each round r by ( CK^{(r)} = f_5({CK_j}) ). We argue that, under majority-honest assumptions and token-weighted incentives, honest play is a Nash equilibrium and the framework converges to a unique fixed point even in Byzantine settings.

While first applied to federated learning, the framework is underline{not} limited to distributed ML training. It supports socially scoped cognition, where actors contribute local knowledge and reasoning to a shared space, enriching common knowledge iteratively.

Two case studies illustrate its range: (i) emph{crowd-sourced RLHF}, where dispersed annotators guide policy updates without sharing labels; (ii) emph{federated LoRA tuning} of large language models across siloed hospitals, achieving near-centralized accuracy under HIPAA and GDPR.

We place the framework within an emerging “AI-native stack”: a Byzantine ledger for immutable state, Google’s A2A for agent-to-agent messaging, and Anthropic’s MCP for secure tool access — together enabling a emph{Cognitive Internet} where autonomous agents learn, trade, and verify knowledge without centralized trust, creating a scalable, interoperable substrate for next-generation human–AI collaboration.

Keywords: Neural Networks and their Applications, Deep Learning, Application of Artificial Intelligence
Abstract: This study conducts a comprehensive comparison of four neural network architectures: Convolutional Neural Network, Capsule Network, Convolutional Kolmogorov–Arnold Network, and the newly proposed Capsule-Convolutional Kolmogorov–Arnold Network. The proposed Capsule-ConvKAN architecture combines the dynamic routing and spatial hierarchy capabilities of Capsule Network with the flexible and interpretable function approximation of Convolutional Kolmogorov–Arnold Networks. This novel hybrid model was developed to improve feature representation and classification accuracy, particularly in challenging real-world biomedical image data. The architectures were evaluated on a histopathological image dataset, where Capsule-ConvKAN achieved the highest classification performance with an accuracy of 91.21%. The results demonstrate the potential of the newly introduced Capsule-ConvKAN in capturing spatial patterns, managing complex features, and addressing the limitations of traditional convolutional models in medical image classification.

Keywords: Computational Intelligence, Application of Artificial Intelligence, Deep Learning
Abstract: In recent years, reinforcement learning has made significant progress in a wide range of domains, including autonomous driving, behavior analysis in electricity markets, and robotic control. Many of these studies have demonstrated the effectiveness of applying Multi-Agent Systems (MAS), where multiple agents act cooperatively or competitively. However, a major challenge in MAS lies in the increased environmental uncertainty caused by the influence of other agents, which can lead to instability in the learning process. To address this issue, various approaches have been proposed that aim to improve learning efficiency by leveraging the experience data of other agents. Previous studies have reported that sharing full experience data among agents can enhance learning efficiency in both cooperative and competitive settings. Furthermore, there are cases where sharing only partial experience data has also led to performance improvements. In this paper, we focus on the similarity between agents as a key criterion for selecting shared data and demonstrate its effectiveness in improving learning performance in MAS. Specifically, we propose a novel method that selectively shares experience data based on agent similarity and utilizes this data to complement an agent’s own experiences. Through simulations with heterogeneous (asymmetric) agents, we show that the proposed method enhances learning efficiency in multi-agent environments.

Keywords: Evolutionary Computation, Metaheuristic Algorithms, Swarm Intelligence
Abstract: The growing integration of renewable energy sources (RESs) within electric distribution systems poses substantial challenges, such as network constraint violations and increased monetary costs. Distribution system reconfiguration is a viable solution for addressing these problems. However, due to severe network constraints, distribution system operators often face challenges in finding optimal reconfiguration solutions within practical computational time. This paper proposes a constrained multiobjective evolutionary algorithm (CMOEA) specialized for solving the reconfiguration problems based on dual-stage and dual-population strategies to tackle these difficulties. The proposed CMOEA employs specialized selection pressures for each population in each stage to guide the populations toward good objective values with diversity. The proposed CMOEA is applied to computational experiments of a reconfiguration problem characterized by severe constraints and uncertainties arising from RESs. The results demonstrate its effectiveness in identifying feasible solutions with good objective values, i.e., the low investment cost.

Keywords: Application of Artificial Intelligence, Expert and Knowledge-Based Systems, Evolutionary Computation
Abstract: This paper proposes LLM-GA, a novel framework that integrates large language models (LLMs) with genetic algorithms (GA) for automated trading strategy generation. The system architecture comprises three synergistic modules: 1) a signal generator extracting technical, fundamental, and sentiment indicators; 2) an LLM-enhanced GA core that initializes seed strategies and performs semantically-aware crossover/mutation operations; and 3) an execution module forming a closed-loop adaptive system. Unlike traditional GA that randomly combines signals, our approach leverages LLMs' financial reasoning capability to maintain logical consistency during strategy evolution. Experiments based on historical data of the Chinese stock market in the past five years (2020-2024) show that, LLM-GA achieves superior risk-adjusted returns (Annualized Excess Return (AER)=12.3%, Maximum Drawdown (MDD)=35.2%) compared to baseline methods including vanilla GA, PSO, and ensemble learning. Ablation studies reveal that LLM-guided initialization improves starting strategy quality by 215%, while semantic crossover reduces invalid strategies by 83.5%. Despite performance gaps against RL methods (2-3% lower AER), our method provides unique advantages in strategy interpretability and diversity, addressing critical limitations in black-box approaches like reinforcement learning. The work establishes a new paradigm for human-AI collaborative quantitative strategy development.

Keywords: Fuzzy Systems and their applications, Hybrid Models of Neural Networks, Fuzzy Systems, and Evolutionary Computing, Neural Networks and their Applications
Abstract: The peg transfer task is a crucial exercise in the Fundamentals of Laparoscopic Surgery (FLS) program for developing the dexterity and coordination skills required for laparoscopic surgery. It involves a surgeon holding a grasper in each hand to transfer six pegs on a peg board from the left to the right (or vice versa) and then back again. This task must be accomplished using both graspers while adhering to various constraints on their movement and time. Automating the assessment of this task is crucial for providing objective, consistent, and real-time feedback to laparoscopic surgery trainees, helping them refine their skills more effectively. Previous work developed a neural network to identify grasper and peg objects in various states. The identified states were further used as inputs to a two-level, cascaded, fuzzy logic system to assess grasper movements. The system ran entirely on a PC but was too slow to provide real-time feedback. To address this limitation, this paper presents the development of a custom fuzzy grasper assessment hardware accelerator using VHDL and its implementation on an SoC-FPGA. The accelerator successfully performs the same task while being, on average, over 64 times faster than the previous PC-based implementation. It is capable of 9,211 fuzzy inferences per second. In future work, the neural network running on the PC will be accelerated using the same SoC-FPGA, and the accelerator can be further pipelined to improve performance. This will pave the way for fully integrated real-time feedback for the trainees, a promising advancement in laparoscopic surgery training.

Keywords: Fuzzy Systems and their applications, Machine Learning, Neural Networks and their Applications
Abstract: Generating fuzzy rule bases from data is essential for building interpretable fuzzy systems. Traditional approaches like Wang-Mendel (WM) rely on correlations but often produce large, redundant rule bases, reducing interpretability and increasing computational cost. To address this, recent work has incorporated causal discovery into rule generation. Te Zhang et al.introduced a method using directed graphs within the Markov blanket, but their reliance on DirectLiNGAM limits applicability to linear data. This paper adopts the Causal Additive Model with Unobserved Variables (CAMUV) to identify a target variable’s Markov blanket and extract its direct causal features. These are then used as inputs to a Takagi-Sugeno-Kang (TSK) fuzzy system. Compared to causal learning algorithms like GRaSP, DECI, and BOSS, CAMUV better handles nonlinear and partially unobserved data, enhancing causal discovery and interpretability. Unlike WM, the TSK system generates dynamic causal if-then rules, improving both interpretability and modeling power. Experiments across seven datasets show an average accuracy improvement of approximately 5% over benchmark models. This work offers a novel, causality-driven approach to constructing interpretable fuzzy systems for complex data.

Keywords: Artificial Immune Systems, Hybrid Models of Neural Networks, Fuzzy Systems, and Evolutionary Computing, Expert and Knowledge-Based Systems
Abstract: The Dendritic Cell Algorithm (DCA), inspired by the Human Immune System, is a promising Artificial Immune System (AIS) for anomaly detection. However, its pre-processing phase traditionally depends on expert manual intervention, limiting scalability and objectivity. This study investigates the impact of integrating automated feature reduction techniques to streamline this phase. We propose three approaches: Kernel Principal Component Analysis (KPCA), Autoencoders (AE), and a hybrid Autoencoder with KPCA (AEkPCA). The models were tested on seven datasets from cybersecurity, biology, and finance domains. KPCA, particularly with a Gaussian kernel, delivered the most consistent results, achieving high accuracy and strong MCAV separation. AEkPCA showed competitive performance in complex datasets, especially with polynomial kernels, though with greater variability. AE in isolation exhibited unstable behavior and inconsistent detection. These results support the viability of automated preprocessing in DCA, highlighting that performance depends heavily on the feature reduction method and kernel combination.

Keywords: Hybrid Models of Neural Networks, Fuzzy Systems, and Evolutionary Computing, Deep Learning, Machine Learning
Abstract: This paper proposes a graph neural network model, DJ-HO-GCN, which integrates high-order graph convolution with diffusion distance, addressing the performance limitations of traditional graph neural networks on heterogeneous graphs. This model employs a novel approach grounded in the rich mathematical theory of simple complexes (SCs), a robust tool for simulating high-order interactions. By utilizing this method, the high-order convolution module effectively captures correlations between nodes at greater distances, particularly when analyzing with graphs with complex structures and significant heterogeneity, thereby demonstrating clear advantages. Meanwhile, by introducing the concept of diffusion pump, the diffusion distance of K-step nodes is dynamically calculated during the training process, and a structural filter is generated in combination with the diffusion distance. This model can well describe the similarity between nodes, enhance the modeling ability of the model for remote dependencies, and effectively reduce the problem of excessive smoothing. To further improve the applicability of the model, DJ-HO-GCN combines these two mechanisms, avoiding the limitations of traditional methods and demonstrating strong robustness and scalability in different types of graph data. The experimental results show that DG-HO-GCN has achieved a high accuracy rate on most datasets, especially showing obvious advantages on heterogeneic graphs (such as chameleons and squirrels), thereby verifying the effectiveness and innovation of this method in graph neural networks.

Keywords: Hybrid Models of Neural Networks, Fuzzy Systems, and Evolutionary Computing, Evolutionary Computation
Abstract: Neuroevolution of Augmenting Topologies (NEAT) stands out as a pioneering framework for simultaneously evolving weights and architectures of artificial neural networks; however, several implicit design choices remain underexplored, particularly in large-scale or zero-knowledge scenarios. This paper revisits the canonical NEAT algorithm and the widely adopted neat Python library, presenting an extensive experimental campaign that spans synthetic control tasks. serves its dual goals of methodological critique and targeted enhancement. The aim is to introduce lightweight yet impactful extensions, including: balanced crossover, rejuvenation-driven speciation, adaptive hyper-parameters, and GPU-accelerated batch evaluation. Thus, achieving improved convergence speed, solution compactness, and behavioral diversity. Some empirical results demonstrate consistent performance gains across the examined domains, affirming the practical value of our refinements. These contributions advance NEAT toward a more scalable, adaptive, and high-performance solution for automated neural architecture discovery.

Keywords: Haptic Systems, Human Factors, Human-Machine Interaction
Abstract: There is a trend in human-machine interaction towards a more physical interaction between humans and machines. Examples are shared control systems such as lane keeping assistants or human-robot collaboration. Thereby, the interplay of the sensorimotor control loops of the human operator with the machine plays an important role. State of the art literature suggests that this interplay should be considered already in the design process of such systems. Therefore, the realistic modeling of the human motor control is crucial for a human model for physical human-machine interaction. The current state of research in neuromechanics suggests that humans use so-called internal models of their limbs and environment, located in the central nervous system, to perform accurate movements despite slow reaction times. This theory is called the internal model hypothesis. This paper presents an approach to implement the internal model hypothesis in a dynamic human model using inverse models. The inverse models are derived using modified Newton-Euler equations. Finally, the validity of the chosen approach is demonstrated by comparing numerical simulations with experimental data using the example of a human-steering wheel interaction. The numerical simulations show a similar qualitative behaviour compared to the mean subject trajectories. Furthermore, the simulated trajectories are within the standard deviation of the experimental data for most of the time.

Keywords: Haptic Systems, Human-Machine Interaction
Abstract: In rehabilitation of upper limb impairment, grip training is a crucial step and requires a universal approach. To address rehabilitation of somatosensory perception and grip control we present a hand grip assessment and training device with integrated tactile feedback for immersive, interactive therapy. The grip is soft and designed to measures grip force omnidirectionally. Additionally, texture simulation is provided through a high-resolution vibration module, delivering tactile feedback. The mechanical design consists of a base platform and a precisely aligned steel support element that holds a grip measurement unit. The system supports grip measurements and offers a playful, engaging experience designed to enhance patient motivation during therapy by integrating VR game applications. To verify that the proposed device enables omnidirectional grip assessment, we measured the force values by applying the same force at six different angles. Each of the six force values was repeatedly measured across different angles. The device allows users to grip freely, making it suitable for rehabilitation training, and offers significant potential for personalized therapy applications.

Keywords: Haptic Systems, Human-Machine Interaction, Wearable Computing
Abstract: This paper presents the Funabot-Suit for the Upper Body, enabling users to perceive seven upper body motions (shoulder abduction, elbow flexion/extension, and trunk rotation) through kinesthetic feedback induced by 80 McKibben artificial muscles embedded in elastic clothing. Unlike existing systems that rely on rigid structures or vibrotactile feedback, the Funabot-Suit provides soft, multi-joint kinesthetic feedback via garment deformation. Building on our previous work focused on trunk motion, this study extends perception to more localized joints such as the shoulders and elbows. The experimental results from nine participants indicate that larger body parts tend to induce stronger and clearer sensations, while certain localized motions, such as elbow extension, are perceived less clearly. These findings suggest that both the size of the body part and the placement of artificial muscles affect kinesthetic perception. The results provide design insights for future wearable devices targeting VR, rehabilitation, and human-robot interaction.

Keywords: Haptic Systems, Intelligence Interaction, Human-Machine Interaction
Abstract: This paper introduces HapticVLM, a novel multimodal system that integrates vision-language reasoning with deep convolutional networks to enable real-time haptic feedback. HapticVLM leverages a ConvNeXt-based material recognition module to generate robust visual embeddings for accurate identification of object materials, while a state-of-the-art Vision-Language Model (Qwen2-VL-2B-Instruct) infers ambient temperature from environmental cues. The system synthesizes tactile sensations by delivering vibrotactile feedback through speakers and thermal cues via a Peltier module, thereby bridging the gap between visual perception and tactile experience. Experimental evaluations demonstrate an average recognition accuracy of 84.67% across five distinct auditory-tactile patterns and a temperature estimation accuracy of 86.7% based on a tolerance-based evaluation method with an 8°C margin of error across 15 scenarios. Although promising, the current study is limited by the use of a small set of prominent patterns and a modest participant pool. Future work will focus on expanding the range of tactile patterns and increasing user studies to further refine and validate the system's performance. Overall, HapticVLM presents a significant step toward context-aware, multimodal haptic interaction with potential applications in virtual reality, and assistive technologies.

Keywords: Haptic Systems, Virtual and Augmented Reality Systems, Human-Machine Interaction
Abstract: In recent years, kinesthetic illusion has gained attention as a means of presenting a sense of movement to users and improving immersion in human-machine systems for entertainment, skill acquisition training, and rehabilitation. Kinesthetic illusion refers to the phenomenon in which a stationary body part is perceived to move when vibratory stimulation is applied to muscles. However, the intensity of the illusion varies among individuals and some people do not perceive the illusions. To address this issue, this study proposes a method to induce and enhance kinesthetic illusions in the wrist joint by combining finger joint extension with vibratory stimulation. Finger joint extension is closely associated with wrist movements in daily activities. Therefore, by simultaneously presenting finger joint extension with vibratory stimulation, this approach aims to evoke the user's motor imagery and strengthen the illusion. A psychophysical experiment confirmed that the proposed method induced stronger illusions than the vibratory stimulation alone. Furthermore, this study revealed that the angle to which the fingers are extended is a key parameter for enhancing the illusion using the proposed method.

Keywords: Haptic Systems, Virtual/Augmented/Mixed Reality, Virtual and Augmented Reality Systems
Abstract: In virtual reality (VR) environments, enhancing force feedback is crucial for achieving immersive and natural interactions. Force feedback devices are generally used to present stiffness and resistance, allowing users to perceive physical properties of virtual objects. However, even with force feedback, users may fail to perceive contact despite visual confirmation that the object is being touched. To address this issue, we developed a multimodal haptic presentation system that combines force feedback using a magnetorheological (MR) brake with vibrotactile feedback. The integration of vibration aims to complement force feedback, enhancing the sensation of contact. Experiments were conducted to evaluate the effectiveness of the proposed system by comparing four feedback conditions: no feedback, force feedback only, vibrotactile feedback only, and combined feedback. The results indicated that the combination of force and vibrotactile feedback significantly improved contact perception and pressing sensation, especially in upward and forward pressing tasks. Additionally, vibrotactile feedback alone was found to enhance contact sensation compared to force feedback alone, suggesting that vibration effectively compensates for contact perception when force feedback is insufficient. These findings demonstrate that combining force and vibrotactile feedback contributes to more realistic and intuitive haptic experiences in VR, providing valuable insights for designing advanced haptic interfaces.

Keywords: Electric Vehicles and Electric Vehicle Supply Equipment, Intelligent Transportation Systems, Large-Scale System of Systems
Abstract: This study developed an application-specific integrated circuit (ASIC) with a speed controller, a nine-stage error fuzzy controller, and a discrete multiple vector voltage (DMVV) system for modified direct torque control (MDTC). By using the nine-stage error fuzzy controller, the proposed system effectively stabilizes a motor’s flux and ensures high control precision by incorporating speed feedback. This feature enables the system to ensure that the flux and torque values are close to the designed values, which results in high motor performance. A DMVV switching table plays a crucial role in facilitating the appropriate six-switch signals on the basis of the modified flux and torque signals from the fuzzy controller. The proposed DMVV system considerably reduces ripples and enhances overall system stability by generating more vector voltages than those generated in the conventional DTC method. The proposed system architecture and functional modules were implemented using Verilog hardware description language. After the syntax and functionality of the designed ASIC were rigorously verified using a field-programmable gate array development board, the designed ASIC was fabricated through the 0.18-μm complementary metal–oxide–semiconductor process of Taiwan Semiconductor Manufacturing Company. This ASIC caters to the specific requirements of three-phase induction motors. Measurement results indicated that the fabricated ASIC had a chip area of 0.974 * 0.976 mm2, a sampling frequency of 40 MHz, and power consumption of 0.5957 mW under a supply voltage of 1.8 V and an operating frequency of 10 MHz.

Keywords: Communications, Large-Scale System of Systems, Smart Sensor Networks
Abstract: An 8-tap feed-forward equalizer (FFE) and 10-tap decision feedback equalizer (DFE) were designed for the IEEE 802.3u 100Base-TX specification. The weights of these adaptive filters are updated with a sign–sign least-mean-square (SSLMS) algorithm. To ensure the flexibility of the circuits for various environments channel lengths, the equalizers are designed such that the number of taps can be adjusted to between 1 and 8 for the FFE and 1 and 10 for the DFE. Results indicated that the equalizer can compensate for channel delays of more than 20 dB, and it achieved a bit error rate of 2 × 10−3. A high-speed serializer/deserializer communication application-specific integrated circuit was designed in Verilog Hardware Description Language and implemented on the TSMC 90-nm CMOS 1P9M standard cell process. In simulations, the chip area, delay cycle, and logic gate counts were 845.445  845.445 m2, 12 cycles, and 75 187 gates, respectively. The supplied voltage and frequency of the proposed adaptive DFE were 1.2 V and 250 MHz, respectively.

Keywords: Smart Sensor Networks, System Modeling and Control, Consumer and Industrial Applications
Abstract: Visually monitoring mango leaf plants can increase optimal growth through early detection of leaf disease. However, the high complexity of mango leaves with multiple intersections and noisy background causes the model performance to be less sensitive to the actual condition of the leaf plant. To address this issue, the proposed model will be constructed by integrating YOLOv10 and the improved multi-scale attention module. Specifically, the detection head structure of YOLO will be fused with the attention module which consists of the integrated efficient multi-scale attention (EMA) module and non-local block mechanism. This mechanism not only improves the model performance, especially in long-range dependencies and tiny objects but also provides an adaptable lightweight model for edge computing systems. The experimental results indicate the proposed model has outstanding performance with mAP50 of 0.944. Meanwhile, the original model gains the mAP50 of 0.917. Compared with other models including ablation study, comparative model experiment, and practical application, the proposed model achieves the best overall detection performance.

Keywords: Control of Uncertain Systems
Abstract: This paper presents an approach for the forecast of daily stock price. Stock prices are not constant over time and are becoming increasingly uncertain in modern financial markets, so that their forecasting is more important and challenging. A new structure, called a cerebellar model extreme learning machine (CMELM), is proposed, which includes a cerebellar model neural network used as a main predictor and an extreme learning machine used for parameter learning. In order to attain better accuracy, a wavelet is used to decompose the original stock price time series. This framework is tested using the data from the Taiwanese stock market, and the experimental results show that it outperforms the benchmarks that are established in this study. Because it is extremely fast and sufficiently accurate, the proposed method has great potential for practical applications.

Keywords: Adaptive Systems, Consumer and Industrial Applications, System Modeling and Control
Abstract: Conventional automatic speech recognition (ASR) systems often encounter difficulties in processing speech of dysarthric people, which is characterized by impaired neuromuscular control and atypical articulation. Existing commercial solutions, such as Google Speech to Text and Microsoft Azure Speech, exhibit high error rates when processing the speech of people with dysarthria, severely hindering the accessibility of communication. This study introduces a novel personalized ASR system that aims to address these challenges by integrating speaker adaptive modeling and advanced correction mechanisms. The system is based on the transformer-based SenseVoice mini-model, which utilizes the functionality of the transformer-based Large Language Model (LLM) to extract speaker-specific speech features from historical data and implement a dynamic history correction algorithm. By employing an LLM-based style extraction method, the system develops a personalized correction model that can be adapted to individual speech variations. Experimental evaluations show that the system provides substantial improvements in recognition accuracy and processing efficiency compared to traditional automatic speech recognition (ASR) techniques. This research highlights the potential of AI-driven personalization approaches in assistive communication technologies, providing a promising avenue for enhancing communication tools for people with speech impairments.

Keywords: Communications, Adaptive Systems, Decision Support Systems
Abstract: We consider the problem of entity alignment in multi-modal knowledge graphs (MMKGs). The explosion in interest in MMKGs has led to a wide array of techniques used to align entities within multiple MMKG’s. The nature of the different modalities available in the MMKG makes the act of embedding them in a shared space challenging. This paper proposes using optimal transport (OT) as a means of embedding and aligning multiple modalities into one unified representation. After acquiring the unified representation, we propose using contrastive learning to train a model for better performance in accurately predicting links between entities. Experiments looking at the effect of the additional modality and OT data fusion show poor performance failed to meet, let alone exceed, current state of the art.

Keywords: Active BMIs, BMI Emerging Applications, Other Neurotechnology and Brain-Related Topics
Abstract: A brain-computer interface (BCI) is a technology that enables direct communication between a user and external devices through brain activity. BCIs can be particularly beneficial for individuals with severe motor impairments who are unable to use conventional assistive technologies that require muscular control. The present study evaluates a preliminary ERP-based BCI designed as an augmentative and alternative communication (AAC) system for this population. Ten able-bodied participants were asked to operate the AAC-BCI system, which enabled stepwise selections through pictogram-based hierarchical menus to address various communication needs. Both objective performance metrics and subjective user feedback were collected. The results validate the feasibility of the proposed system as an AAC-BCI solution. However, further improvements are necessary to optimize its usability and effectiveness for the target population. Future research should focus on refining the system’s performance and adaptability to enhance its practical application in real-world scenarios.

Keywords: Active BMIs, BMI Emerging Applications
Abstract: A brain-computer interface (BCI) typically requires a calibration phase to train its decoding model using supervised data, which can be time-consuming and impractical in real-world scenarios. This study eliminates the need for such calibration by investigating two zero-training approaches using a P300 event-related potential (ERP) dataset. We first evaluate reconvolution canonical correlation analysis (CCA) on ERP data, which marks its first application beyond its original domain of code-modulated visual evoked potentials (c-VEP). Second, we examine unsupervised mean-difference maximization (UMM), a recently proposed method for calibration-free ERP decoding. For both approaches, we assess performance under two conditions: instantaneous classification and cumulative learning across previously classified trials. Surprisingly, despite CCA originating from the c-VEP domain, our results demonstrate that both CCA and UMM achieve comparably high classification accuracy (97 % and 100 %) in decoding ERP data without requiring a calibration session. We highlight the unique advantages that each method offers and discuss implications for future research. By enabling reliable, calibration-free decoding, this work supports the development of more practical and accessible BCIs across various stimulus paradigms and applications. The potential integration of CCA and UMM presents a promising avenue to further enhance BCI usability.

Keywords: BMI Emerging Applications, Active BMIs
Abstract: Brain–machine interfaces (BMI) for lower‑limb exoskeletons are a state‑of‑the‑art neurorehabilitation modality. They decode electroencephalographic (EEG) recordings during motor imagery (MI)—the mental rehearsal of movement—to infer intent and drive exoskeleton control. Yet MI decoding suffers from low signal‑to‑noise ratio, EEG non‑stationarity, and high inter‑trial/subject variability. Conventional machine‑learning classifiers further struggle with limited training data and overfitting, undermining real‑time robustness. In this preliminary, offline study on a single subject, a novel semi‑supervised MI‑classification network is implemented that includes an L2‑normalized autoencoder with dual reconstruction and classification branches—that, to our knowledge, is the first correctly tailored for closed‑loop lower‑limb exoskeleton control. This method is compared against four supervised approaches using a hybrid feature‑extraction pipeline capturing spectral, spatial, and temporal EEG dynamics. Supervised models were evaluated via leave‑one‑out cross‑validation, while the semi‑supervised framework’s latent representations were examined with K‑means clustering and t‑Stochastic Neighbour Embeddings (t-SNE). Event‑based false‑positive (FPR) and true‑positive ratios (TPR) served as comparative metrics. All approaches achieved 61–67 % accuracy, with the semi‑supervised network showing a lower FPR—suggesting its promise for more robust, data‑efficient BMI‑driven exoskeleton control.

Keywords: Other Neurotechnology and Brain-Related Topics, BMI Emerging Applications
Abstract: This paper proposes to investigate the design of user-friendly reactive Brain-Computer Interfaces (rBCIs) based on Code-Modulated Visual Evoked Potentials (c-VEP). The BCI was implemented using the StAR-Burst paradigm, which features small, randomly-oriented texture patches designed to optimize foveal neural responses while enhancing user visual comfort. We extended this paradigm to support an 11-class selection scenario using dry EEG electrodes. In addition, the study explored the impact of predictive visual feedback on user experience. Two feedback types—Halo and Depth—were compared against a control condition with no feedback, aiming to enhance users' sense of control during interaction. Results demonstrated that the BCI achieved high classification accuracy with a dry EEG system across all three conditions (mean = 93,3%), using only 33 seconds of calibration data. Contrary to our expectations, predictive feedback did not lead to significant improvements in classification accuracy or decoding time. However, the Halo feedback significantly increased users' anticipation of success, though it also caused greater peripheral distraction. Interestingly, decoding time improved significantly with practice, underscoring the role of user adaptation in enhancing performance. Overall, the findings highlight the need for explicit training to help users effectively interpret and utilize predictive feedback.

Keywords: Active BMIs, Other Neurotechnology and Brain-Related Topics
Abstract: Real-time brain-computer interfaces (BCIs) require fast and accurate alpha-band power estimation. However, alpha-waves are low-frequency components, which makes their rapid extraction using conventional methods, such as bandpass filters (BPF), difficult. This study applies a nonlinear finite-time observer (FTO) to extract low-frequency features from electroencephalogram signals. Experimental results demonstrate that the proposed FTO achieves a convergence time of 4 ms compared to 700 ms with BPF. Moreover, the FTO captures -8 dB attenuation as motor imagery-related alpha-band power, twice the magnitude obtained with conventional methods (-4 dB). Thus, compared to conventional approaches, the FTO enhances both responsiveness and clarity in detecting motor imagery-related reductions in alpha-band.

Keywords: Active BMIs, BMI Emerging Applications, Other Neurotechnology and Brain-Related Topics
Abstract: A Brain–computer interface (BCI) enables direct control of external devices through neural activity. Among BCI paradigms, visual evoked potentials (VEPs) are widely used because they harness the brain’s response to visual stimuli to deliver an intuitive and inclusive control interface. However, these systems typically rely on low-frequency flickering stimuli to evoke strong neural responses, which can cause discomfort and fatigue. While previous attempts have sought to mitigate this discomfort, eliminating it entirely would be ideal. This could be achieved with high-frequency flickering operating above the critical fusion frequency threshold, which is invisible to the human eye yet still evokes a distinct neural imprint. While this neural response is distinguishable from that of a non-flickering stimulus, reliably differentiating between two different high-frequency stimuli remains challenging. Our novel Imperceptible VEP (I-VEP) paradigm combines imperceptible high-frequency flicker segments with static (non-flickering) intervals to create patterns analogous to steady-state VEP (SSVEP) and code-modulated VEP (cVEP). From this paradigm, we define two variants: Imperceptible Steady-State VEP (I-SSVEP) and Imperceptible Code-Modulated VEP (I-cVEP). Our initial I-cVEP experiments demonstrate its feasibility, achieving a mean bit-wise accuracy of over 92% across 27 participants. These findings may represent a significant first step toward fully comfortable visual-stimulus BCIs, enabling broader adoption and improved usability.

Keywords: Big Data Computing,
Abstract: John Krumm graduated from the School of Computer Science at Carnegie Mellon University in 1993 with a PhD in robotics and a thesis on texture analysis in images. He worked at the Robotics Center of Sandia National Laboratories in Albuquerque, New Mexico for the next four years. His main projects there were computer vision for object recognition for use in robots and vehicles. He was at Microsoft Research in Redmond, Washington, USA for 25 years, starting in 1997. He is currently an associate director of the Integrated Media Systems Center in the Viterbi School of Engineering at the University of Southern California. His research focuses on understanding peoples' location and personal data privacy. In 2017 he received a 10-year impact award for a paper on location privacy from the ACM UbiComp conference, and another from the same conference in 2021. He received the best paper award at the ACM SIGSPATIAL conference in 2022 and at the Mobile Data Management conference in the same year. His h-index on Google Scholar is 75. He is an inventor on 82 U.S. patents. Dr. Krumm was a PC chair for UbiComp 2007, ACM SIGSPATIAL 2013, and ACM SIGSPATIAL 2014. He is a past coeditor in chief of the Journal of Location Based Services and past associate editor for ACM Transactions on Spatial Algorithms and Systems. He currently serves on the editorial board of IEEE Pervasive Computing Magazine. He is the chair of the executive committee of ACM SIGSPATIAL and part of the Science Advisory Committee of the Geospatial Science and Human Security Division at Oak Ridge National Laboratory. He is an editorial fellow for the Paris Institute for Advanced Study.

Keywords: Deep Learning, Image Processing and Pattern Recognition, AI and Applications
Abstract: Alzheimer's disease (AD) is the leading cause of dementia in the elderly, with early diagnosis being a primary goal. Methods based on convolutional neural networks and attention, such as the LA-GMF model, offer strong performance and interpretability. We present a LA-GMF-based framework aimed at improving transparency and reliability. After retrieving the optimal number of attention heads through experimentations with a dataset from ADNI, for cross-validation, and the MIRIAD dataset as a hold-out set, we introduced a confidence threshold to reject uncertain predictions. Our proposed framework achieved 99.38% accuracy, and 98.97% sensitivity on accepted samples, outperforming the original LA-GMF while retaining 75.7% of the test set, enhancing system transparency to potentially achieve stakeholder trust.

Keywords: Deep Learning, Image Processing and Pattern Recognition, Application of Artificial Intelligence
Abstract: Although Transformer has been successfully applied to Vision tasks in various fields, its large computational cost and performance degradation due to divergence from the language task are issues to be addressed. In this paper, we introduce token pruning to Mask2Former, a state-of-the-art segmentation method, to reduce computational cost and improve recognition accuracy without additional training. Multi-Scale Token Pruning (MSTP) works effectively on the multi-scale feature tokens of Mask2Former and can be universally implemented with various conventional token pruning methods. Experimental results show that introducing Top-K (norm+rand) MSTP into the Mask2Former of Swin-L backbone achieves +0.28 (56.31) mIoU on the ADE20K benchmark with +5.7% speed up. With this improvement, Mask2Former+MSTP can achieve mIoU equivalent to the large and powerful BEiT-UperNet with 1/4 of the computational complexity. In addition, +0.04 (57.86) PQ for COCO panoptic and +0.19 (63.36) mIoU for Mapillary Vistas are achieved, showing particular usefulness in complex semantic tasks with a large number of categories.

Keywords: Deep Learning, Image Processing and Pattern Recognition, Machine Learning
Abstract: User identification through aesthetic preferences has gained attention as a promising direction in social behavioral biometrics, offering a non-invasive and privacy-conscious alternative to traditional physiological identifiers. Unlike still images or single-modal inputs, video-based aesthetic preferences provide richer, temporally-aware insights into user behavior, enabling a deeper understanding of personal taste and style. Despite this potential, existing approaches often treat preference items independently and fail to capture the interrelationships within a user’s preference set. To address these limitations, this paper introduces Preference-Aware Set Encoding with Contrastive Personalization (PASE), a novel deep learning framework designed to model user identity based on structured preferred video sets. The proposed method integrates a Cross-Video Attention Encoder to learn co-preference patterns across videos, a User-Anchor Contrastive Loss to align personalized embeddings, and a Cross-Set Mixup Regularization technique to improve generalization by simulating diverse preference scenarios. Evaluation on a curated aesthetic dataset demonstrates that PASE achieves 98.38% identification accuracy, outperforming unimodal and multi-modal baselines. These findings highlight the unique advantages of leveraging video-based aesthetic information for biometric identification, particularly in applications demanding both accuracy and user-friendly privacy safeguards.

Keywords: Deep Learning, Neural Networks and their Applications, Image Processing and Pattern Recognition
Abstract: In recent years, several advances have been observed in Deep Learning with surprising results. Models in this area have been increasingly used in numerous applications, including those sensitive to human life, which require clear explanations and justifications. This has encouraged several types of research into the explainability of neural networks. Various explainability methods have been proposed, but not many metrics to evaluate these methods. The most commonly used metric is the Intersection over Union (IoU). It is applied between two bounding boxes to allow one to measure the similarity between them. However, due to the characteristics of the results of the explainability methods, called saliency maps, which do not have a known shape, we hypothesise that there must be a better metric that allows one to find an explainability method that produces results that best resemble the human perception. This work proposes using different metrics to assess the similarity between human perception and the explanation saliency maps to find a better metric. To this end, an investigation was conducted employing a subset of the ImageNet dataset, which corresponded to the Chihuahuas images. Several CAM-based explainability methods were used to generate saliency maps, which were compared with human perception of the most important parts of the chihuahuas in each image. Alignment was measured by applying distance metrics between the bounding box of human annotations and the saliency maps produced by each explainability method. Rankings of the best saliency maps were created using the results of the distance metrics and compared to the ranking obtained using people's choice, collected through crowdsourcing, of the best explanation saliency maps for each selected image. Comparison between rankings was performed using the Rank-Biased Overlap (RBO) metric. The results indicate the feasibility of our method to find the explainability method that best resembles human perception. In our experiments, the two metrics that best resemble human perception corresponded to Manhattan and Correlation. Besides, the best explainability methods regarding human perception were LayerCAM, Score-CAM, and IS-CAM.

Keywords: Deep Learning, Representation Learning, Transfer Learning
Abstract: Deep learning models for edge deployments must be small, efficient, and robust. Knowledge distillation from large foundation models (FMs) can help, as they capture rich, transferable representations from large, multimodal datasets. However, two key challenges hinder this process: (1) extreme model compression based on a test dataset often leads to brittleness under distribution shifts, and (2) the distribution gap between an FM’s training data and a small model’s target dataset makes standard distillation methods ineffective.

We propose a tunable dynamic loss curriculum for knowledge distillation to address these issues. Experiments show that small encoders trained with our approach achieve balanced transfer learning performance across both primary and out-of-distribution tasks. For instance, a tiny GPT-like model effectively transfers to sentiment classification and language modeling. Likewise, a tiny ResNet trained on CIFAR-10 achieves 10% higher accuracy on corrupted CIFAR-10 than state-of-the-art baselines for tiny models. While it trails dedicated robustness training by 4%, our method ensures superior adaptability across diverse datasets and tasks.

Keywords: Cloud, IoT, and Robotics Integration, AI and Applications, Deep Learning
Abstract: Multi-axis servo coordinated control enables multiple axes to track distinct target trajectories simultaneously. Through networked collaboration, these axes together can achieve complex tasks with enhanced adaptability and flexibility. In this paper, we introduce an edge-cloud collaborative multi-axis servo coordination control framework, exploiting both advantages of edge and cloud computing for optimizing the coordinated control performance of multi-axis servo system. Considering that axis states and control signal sequences are transmitted over a limited shared wireless channel, we formulate a long-term combinatorial decision problem under stringent communication resource constraints. A novel reinforcement Q-knapsack approach is proposed, which solves a grouped knapsack problem at each time step concerning the number of consumed slots and action Q-values, while deep reinforcement learning is utilized to optimize the action Q-value estimation in the long run. Simulation experiments demonstrate that the proposed approach is not only effective but also superior compared to counterparts.

Keywords: Application of Artificial Intelligence, Deep Learning, Machine Vision
Abstract: Fire detection faces challenges in complex scenarios (e.g., industrial plants, outdoor environments, and urban buildings), including blurred features of small targets, background interference, and shape variability. It also lacks support for false alarm verification and situation analysis, making it difficult to meet requirements for detection accuracy and emergency response. To address these issues, this paper proposes a two-stage fire detection and analysis method based on the collaboration between large and small models, aiming to achieve end-to-end optimization from efficient detection to intelligent emergency response. In the first stage, we propose MYGA-YOLO, a fire detection model based on YOLOv11n. It incorporates RFAConv to dynamically adapt to shape variability, the EMA attention mechanism to suppress background interference, and RepGFPN for efficient multi-scale feature fusion to enhance small target detection. These improvements significantly boost detection accuracy and robustness. Experimental results show that its accuracy, recall, and mAP@50 are improved by 3.7%, 2.9%, and 3.5%, respectively, compared to YOLOv11n. In the second stage, a multimodal large language model combined with Retrieval-Augmented Generation technology is employed. Through cross-modal reasoning, it integrates images, sensor data, and fire-fighting knowledge to perform false alarm verification, fire situation analysis, and generate emergency response strategies. By leveraging the collaboration between the computational efficiency of small models and the intelligent reasoning capabilities of large models, this approach addresses the limitations of existing methods in terms of detection accuracy and fire emergency support, offering an innovative solution for fire detection and emergency response in complex scenarios.

Keywords: Application of Artificial Intelligence, Deep Learning, Multimedia Computation
Abstract: Image-text retrieval (ITR) is a fundamental task in multimodal learning. It aims to bridge image and text by constructing a shared embedding space that achieves accurate semantic alignment across two modalities. Most existing work has been devoted to designing tailored cross-attention modules to pursue retrieval accuracy but ignores the learning potential of the model network architecture. This work introduces a dual encoder that uses global information to enhance the feature interaction between visual regions. Designing an Adaptive Pooling (AP) module enables the model to automatically learn the optimal aggregation strategy from multiple perspectives based on the local features. We further propose a Dynamic Triplet Loss (DTL) to adjust the learning objective of the model network dynamically for efficient training. Experimental results on two benchmark datasets, Flickr30K and MS-COCO, demonstrate our APDTL model achieves state-of-the-art performance.Our code is released at github.com/jinshuaihu/APDTL.

Keywords: Application of Artificial Intelligence, Deep Learning, Neural Networks and their Applications
Abstract: Precise detection of eggs on conveyor belts in industrial automated production lines is of significant importance for reducing detection error rates and improving production efficiency. However, existing detection models face considerable challenges in detection accuracy and real-time processing when handling practical scenarios such as densely arranged eggs, complex background interference, and high-speed conveyor belts. This paper proposes a lightweight real-time detection Transformer model called RCD-DETR, consisting of three key innovative components: (1) a lightweight Reparam Context-aware Network (RCNet) that effectively balances feature extraction capability and computational efficiency; (2) a Context-Sensitive Refinement Feature Pyramid Network (CSRFPN) with enhanced contextual awareness and multi-scale feature representation capabilities, strengthening the model's ability to recognize small and dense objects; and (3) a Dilated Reparam Bottleneck C3 (DRepC3) module that expands the receptive field range while further reducing computational resource requirements. Experimental evaluation indicates that, compared to the RT-DETR baseline model, the proposed method achieves improved detection accuracy while reducing computational complexity by 54.1%, decreasing parameter count by 51.3%, and compressing model size by 50.8%. The method achieves an excellent balance between accuracy, inference speed, and resource consumption, making it suitable for deployment on resource-constrained edge computing devices for precise real-time egg detection on conveyor belts.

Keywords: Application of Artificial Intelligence, Deep Learning, Image Processing and Pattern Recognition
Abstract: Kidney stones significantly impact healthcare systems, with diagnosis typically requiring time-consuming Computed Tomography (CT) scan consultations between physicians and radiologists, often delaying patient care. Achieving a quick and accurate diagnosis is essential to ensure timely and effective treatment, which has motivated the development of Deep Neural Network (DNN)-based approaches for automated kidney stone detection. However, building effective models remains challenging, as it often requires access to large and diverse datasets that are siloed across institutions, and sharing such medical data is rarely feasible due to strict privacy regulations and patient confidentiality concerns. This paper proposes a privacy-preserving Federated Learning (FL) framework that enables multiple medical institutions to collaboratively train a DNN model without sharing sensitive patient data. Each institution trains a local model on its private dataset, and a centralized trusted server securely aggregates model parameters. We evaluate our approach using abdominal CT scan image datasets from two distinct institutions. Experimental results demonstrate that our proposed model achieves high classification accuracy within the same training environment, with an F1-score of up to 0.94. In addition, in cross-dataset evaluations, our approach outperforms traditional centralized baselines, showing significantly lower performance degradation while preserving patient privacy.

Keywords: Application of Artificial Intelligence, Deep Learning, Neural Networks and their Applications
Abstract: Multivariate time series forecasting presents a significant challenge across various fields, requiring accurate predictions of future values based on multiple interrelated time series. Recent research has shown that the Channel Independent (CI) approach, which processes each sequence independently, can improve prediction accuracy, but neglecting the relationships between sequences may result in inadequate generalization. Channel Dependent (CD) methods, while integrating all sequences information, however, may compromise prediction accuracy by mixing potentially unrelated data. In this paper, we propose a novel framework called MDF-Mamba which employs a multi-scale decomposition strategy, utilizing the CI approach at fine scales to capture the unique characteristics of individual sequences, thereby enhancing model robustness. At coarse scales, the CD approach is used to capture correlations between sequences, improving the model's generalization capabilities. MDF-Mamba fully considers the individual characteristics of sequences and their interrelationships, balancing Channel Independent and Dependent to improve multivariate time series forecasting performance. Extensive experimental results across multiple real-world time series datasets demonstrate that MDF-Mamba achieves state-of-the-art performance.

Keywords: Robotic Systems, Mechatronics
Abstract: Current training and development platforms in robotics assisted surgery often lack an affordable vision system. In this paper we present a cost-effective stereoscopic camera system for surgical training. Our approach utilizes a double parallelogram mechanism to achieve an approximate remote center of motion with three degrees of freedom: two revolute joints for orbital movement and one prismatic joint for depth positioning. The system incorporates two cameras with integrated illumination to provide depth perception. It can be steered hands-free using a head-mounted display, demonstrating the development potential the platform provides. A proof-of-concept evaluation focused on the control system, demonstrating the system’s ability to achieve precise positioning across all three axes with errors below 1° error for revolute joints and 1 mm error for prismatic joint. As a desktop-compatible and modular platform, we hope that this system significantly reduces barriers to surgical robotics in research, development and training.

Keywords: Robotic Systems, Modeling of Autonomous Systems, Mechatronics
Abstract: The COVID-19 pandemic has severely affected public health, healthcare systems, and daily life, especially amid resource shortages and limited workers. This crisis has underscored the urgent need for automation in hospital environments, particularly disinfection, which is crucial to controlling virus transmission and improving the safety of healthcare personnel and patients. Ultraviolet (UV) light disinfection, known for its high efficiency, has been widely adopted in hospital settings. However, most existing research focuses on maximizing UV coverage while paying little attention to the impact of human activity on virus distribution. To address this issue, we propose a mobile robotic system for UV disinfection focusing on the virus hotspot. The system prioritizes disinfection in high-risk areas and employs an approach for optimized UV dosage to ensure that all surfaces receive an adequate level of UV exposure while significantly reducing disinfection time. It not only improves disinfection efficiency, but also minimizes unnecessary exposure in low-risk areas. In two representative hospital scenarios, our method achieves desired disinfection effectiveness while reducing the time spent evidently. The video of the experiment is available at: https://youtu.be/wHcWzOcoMPM.

Keywords: Robotic Systems, Modeling of Autonomous Systems, System Modeling and Control
Abstract: As an important branch of embodied artificial intelligence, mobile manipulators are increasingly applied in intelligent services, but their redundant degrees of freedom also limit efficient motion planning in cluttered environments. To address this issue, this paper proposes a hybrid learning and optimization framework for reactive whole-body motion planning of mobile manipulators. We develop the Bayesian distributional soft actor-critic (Bayes-DSAC) algorithm to improve the quality of value estimation and the convergence performance of the learning. Additionally, we use a quadratic programming method to calculate and constrain joint velocities, thereby improving the safety of the whole-body motion planning. We conduct experiments and make comparison with standard benchmark. The experimental results verify that our proposed framework significantly improves the efficiency of reactive whole-body motion planning, reduces the planning time, and improves the success rate of motion planning. Additionally, the proposed reinforcement learning method ensures a rapid learning process in the whole-body planning task. The novel framework allows mobile manipulators to adapt to complex environments more safely and efficiently.

Keywords: Robotic Systems, System Architecture, Autonomous Vehicle
Abstract: This paper introduces an extension to the arbitration graph framework designed to enhance the safety and robustness of autonomous systems in complex, dynamic environments. Building on the flexibility and scalability of arbitration graphs, the proposed method incorporates a verification step and structured fallback layers in the decision-making process. This ensures that only verified and safe commands are executed while enabling graceful degradation in the presence of unexpected faults or bugs. The approach is demonstrated using a Pac-Man simulation and further validated in the context of autonomous driving, where it shows significant reductions in accident risk and improvements in overall system safety. The bottom-up design of arbitration graphs allows for an incremental integration of new behavior components. The extension presented in this work enables the integration of experimental or immature behavior components while maintaining system safety by clearly and precisely defining the conditions under which behaviors are considered safe. The proposed method is implemented as a ready to use header-only C++ library, published under the MIT License. Together with the Pac-Man demo, it is available at github.com/KIT-MRT/arbitration_graphs.

Keywords: Robotic Systems, System Modeling and Control, Mechatronics
Abstract: A 3D object grasping point learning system is proposed in this paper, which contains object coordinate construction and grasping point learning. A RGBD Convolutional Neural Network (RGBD-CNN) is proposed to classify the orientation type of objects. An object model and the iterative closest point algorithm (ICP) are then applied to estimate the object pose. Hence, the object coordinate can be constructed in the end. For learning object grasping region, the normal vector images and depth image of the object are obtained first. Then, the grasping range of the end effector (palm) will be simulated on these images. Finally, Convolutional Autoencoder (CAE) is applied to encode the physical characteristics of the simulated palm image. By comparing the features of the simulated palm in the database through a 3D KD-tree, the proposed method can evaluate the grasping points. Through integrating object coordinates and learnt grasping points, the robot plans a suitable grasping point based on the appointed task. It is worth mentioning that most of the research only puts emphasis on either object orientation justification or grasp points generation. However, this research considers the problem of object pose estimation and grasping points generation together. Therefore, the robot can adapt to different task situations in real time. Real experiments demonstrate that the proposed method can recognize various kinds of shapes.

Keywords: Human-Machine Interaction, Haptic Systems
Abstract: Haptic feedback is proposed in sewing machine operations with the objective of reducing errors in seam allowance. Similar to the lane departure warning system in modern cars, such modality of feedback is expected to alert the operator in real time and improve the performance of the human control system. A traditional sewing machine was outfitted with a camera, a single board computer, and an eccentric motor. Experimental results demonstrate the potential for deployment of such systems in apparel manufacturing industries and provide opportunities for greater inclusion of people with disabilities in the workforce.

Keywords: Human-Machine Interaction, Human Factors
Abstract: The shift from driver to passenger may increase the risk of Motion Sickness (MS) for Automated Vehicle (AV) occupants. Motion anticipation, which is believed to mitigate MS, relies to a large extent on visual cues. Yet, the mechanisms through which visual information influences MS remain poorly understood. This paper investigates the effect of visual translation on MS in AVs. In a simulator study, eighteen participants experienced three ‘AV rides’ with identical repetitive braking and accelerating motion on a straight road, differing only in their ‘out-the-window’ view to manipulate the amount of global optic flow. A rural, low optic flow, ride and an urban, high optic flow, ride were compared to a baseline without visual movement. Results suggest that visual translation in both the central and peripheral view, that is congruent with inertial cues, may only slightly reduce MS, as MS seemed to be slightly less in the rural and urban rides compared to the baseline. The amount of global optic flow seems to have little effect, with minimal differences in MS between the rural and urban rides. Nonetheless, it remains uncertain to what extent the type of visual content has affected MS development. A study using more generic visuals could help isolate these effects, by eliminating any recognizable visual elements, while still manipulating the global optic flow rate.

Keywords: Human-Machine Interaction, Human Factors, Assistive Technology
Abstract: Lower-limb exosuits are particularly relevant for individuals with some degree of mobility impairment, such as post-stroke patients or older adults with reduced movement capabilities. This study aims to investigate the mental workload (MWL) assessment of XoSoft, a lower-limb soft exoskeleton, using and comparing subjective and objective physiological metrics. The NASA-TLX questionnaire, the average percentage change in pupil size (APCPS), and the Baevsky stress index (SI) are compared. The experiments were conducted on 18 healthy subjects while walking and involved mathematical tasks to create a double-task condition. The results show a complex interaction between task difficulty, exoskeleton activation, and pupillary dynamics, suggesting that the subject might reach a saturated condition under a high mental load. Besides, the data indicate that pupil diameter may be an objective mental workload indicator that correlates with subjective NASA-TLX questionnaires. The discordant indications from the stress index suggest how different metrics of the ocular and cardiac levels respond differently to various stimuli and dynamics. Research has also revealed ocular asymmetry, with the right eye more sensitive to cognitive load.

Keywords: Human-Machine Interaction, Human Factors, Virtual/Augmented/Mixed Reality
Abstract: Human–autonomy teaming in the Low Altitude Economy (LAE) requires operators to manage both ground and aerial autonomous agents under time pressure, spatial uncertainty, and cognitive load. This study investigates how visual and haptic feedback affect operator situational awareness (SA) in simulated collision avoidance tasks involving cars and drones. A high-fidelity virtual environment was built using Unreal Engine 4 and AirSim, with haptic cues delivered through a wearable bHaptics vest. Twenty-two participants performed within-subject trials across visual-only and visual–haptic conditions. Results showed that haptic feedback significantly enhanced SA, particularly in dimensions related to information acquisition and spare mental capacity. Improvements were more consistent in car-based tasks, while drone scenarios exhibited greater inter-individual variability. These findings demonstrate the potential of multimodal interfaces to support cognitive performance and reduce platform-related disparities in operator SA. This work provides empirical evidence for designing adaptive, perception-aware interfaces in safety-critical human–autonomy teaming systems.

Keywords: Human-Machine Interaction, Human-Computer Interaction, Brain-Computer Interfaces
Abstract: The vision language action (VLA) model has shown extraordinary ability in the motion planning of the robots, which has great value in human-machine-environment interaction (HMEI), especially for rehabilitation engineering and robotic arm teleoperation fields. This paper proposes an HMEI system that integrates the brain-computer interface (BCI) with VLA model, enabling robots to interact with surrounding environment according to human intentions. Firstly, neural decoding technique is employed to interpret neural signals into intentions via a time and frequency fusion model. Then, the user's intentions are converted into a textual instruction, which subsequently guides the VLA model in generating corresponding motion trajectories for robotic arm teleoperation. Experimental results demonstrate smooth and compliant control of the robotic arm, which can be a viable solution for HMEI applications.

Keywords: Human-Machine Interaction, Human-Computer Interaction, Human-Machine Cooperation and Systems
Abstract: This paper proposes an Intelligent Evaluation Agent (IEA) with knowledge graph construction based on the Large Language Model (LLM) and Trustworthy AI Dialogue Engine (TAIDE) for personalized Human-Machine Interactive Learning (HMIL). The intelligent agent will deal with multitasks such as learning data preparation and the learner’s data generation, preprocessing, analysis, and evaluation. Multi-modal data is collected from human-machine interactive activities and processed by an IEA to generate structured data stored in human learning repositories. The intelligent agent focuses on various temporal learning periods, such as macro, meso, and micro-level assessments by integrating Human Intelligence (HI) and Machine Intelligence (MI) results, with the MI-based Genetic Algorithm and Neural Network (GANN) learning mechanism employed to optimize the intelligent evaluation model. The learning data evaluation phase aims to identify a model that best fits the group’s learning behavior through HI-based evaluation and to train it further using MI, ensuring that the trained GANN-IEA model closely approximates the HI-based model. An LLM-based knowledge graph agent also supports the evaluation process by helping teachers analyze and visualize students’ learning progress. Experimental results demonstrate that students who study diligently gain knowledge and exhibit increased interest in learning through HMIL. However, the evidence also suggests that some students who excessively rely on Generative AI (GAI) to reproduce learning content without modification become less inclined to engage in diligent study. Additionally, the proposed IEA effectively reduces teachers’ workload in assessing students’ learning status at the end of the semester and supports personalized learning through the designed HMIL model.

Keywords: System Modeling and Control, Decision Support Systems, Large-Scale System of Systems
Abstract: Fine-grained bird sighting prediction is crucial for advancing ecological research, informing conservation planning, and enhancing the birdwatching experience while fostering public awareness of biodiversity. The rapid expansion of citizen-based bird observation networks has led to an exponential accumulation of bird sighting records, which can be leveraged to train machine learning models for more precise predictions. However, general-purpose machine learning models often fail to incorporate ecological factors that influence bird activity, resulting in less accurate predictions. In this paper, we present an ecologically informed machine learning framework based on LightGBM that integrates spatiotemporal correlations, ecological context, and dynamic environmental variables to improve bird sighting predictions. The framework captures temporal trends using rolling windows, applies spatial smoothing to account for observation proximity, and models ecological dependencies—such as temperature-food interactions—through interaction terms. Key environmental factors, including habitat classifications, weather conditions, and seasonally adjusted food availability proxies, are dynamically incorporated to enhance ecological relevance. Evaluation results demonstrate significant improvements in predictive accuracy, with increased F1-scores compared to baseline methods. By embedding ecological principles into machine learning models, this framework enables data-driven insights that reflect real-world environmental complexities, providing a powerful tool for biodiversity monitoring and conservation strategies.

Keywords: System Modeling and Control, Distributed Intelligent Systems, Discrete Event Systems
Abstract: Sidechains offer partial solutions to Ethereum's scalability challenges; however, they introduce trade-offs related to security and implementation complexity. These limitations have been further addressed by Layer-2 solutions known as rollups, which combine off-chain computation with on-chain verification, preserving both security and decentralization on the Ethereum platform. This paper proposes a Stochastic Petri Net model to evaluate the feasibility of ZK-Rollups by analyzing their impact on throughput and latency. The results indicate that increased adoption of Layer-2 transactions can enhance system throughput by up to 20%. Conversely, latency may rise by more than 100% when larger batches are used, revealing a fundamental performance trade-off.

Keywords: System Modeling and Control, Infrastructure Systems and Services, Communications
Abstract: By locating computational and storage resources at the edge of the network, the emerging paradigm of Mobile Edge Computing (MEC) yields a significant enhancement in user Quality of Experience (QoE). However, the limited resources at edge nodes, coupled with the dynamism of user requests, present a considerable challenge to decision-making in service placement and computational resource assignment. This paper investigates the resource management problem within an edge-cloud cooperative network. Aiming to minimize long-term network latency, the original problem is first modeled as a Markov Decision Process (MDP) featuring a hybrid discrete-continuous action space. To address dynamically arriving tasks and varying network conditions, we develop a Dynamic Service Placement and Computation Resource Allocation (DSPCRA) scheme based on deep reinforcement learning (DRL). DSPCRA integrates a deep deterministic policy gradient (DDPG) with a parameterized action mechanism for online decision-making. Numerous simulations confirm that the proposed scheme exhibits good convergence properties and achieves lower latency performance compared to the benchmark algorithms.

Keywords: System Modeling and Control, Large-Scale System of Systems
Abstract: Modeling and computer simulation are effective tools for predicting system behavior. However, simulation software (i.e., source codes), particularly those on social systems, tends to be complicated and un-reusable owing to the complexity, diversity, and ambiguity of the target systems. This makes it difficult to develop, validate, and explain simulation software. One possible solution is to formalize a mathematically rigorous simulation theory and incorporate it directly into a functional programming language. If the mathematical counterpart of the code is clear, coders can easily explain, validate, and extend it. In this study, we define "monoidal system" as the basis to describe the physical and social system that will be modeled. We also conducted an experiment to demonstrate that the introduction of a monoidal system could reduce program complexity in software development.

Keywords: Decision Support Systems, Consumer and Industrial Applications, Digital Twin
Abstract: Allocation of personnel and material resources is highly sensible in the case of firefighter interventions. This allocation relies on simulations to experiment with various scenarios. The main objective of this allocation is the global optimization of the firefighters response. Data generation is then mandotory to study various scenarios In this study, we propose to compare different data gener- ation methods. Methods such as Random Sampling, Tabular Variational Autoencoders, standard Generative Adversarial Networks, Conditional Tabular Generative Adversarial Net- works and Diffusion Probabilistic Models are examined to ascertain their efficacy in capturing the intricacies of firefighter interventions. Traditional evaluation metrics often fall short in capturing the nuanced requirements of synthetic datasets for real-world scenarios. To address this gap, an evaluation of synthetic data quality is conducted using a combination of domain-specific metrics tailored to the firefighting domain and standard measures such as the Wasserstein distance. Domain-specific metrics include response time distribution, spatial-temporal distribution of interventions, and accidents representation. These metrics are designed to assess data variability, the preservation of fine and complex correlations and anomalies such as event with a very low occurrence, the conformity with the initial statistical distribution and the operational relevance of the synthetic data. The distribution has the particularity of being highly unbalanced, none of the variables following a Gaussian distribution, adding complexity to the data generation process.

Keywords: Biometrics and Applications,, Brain-based Information Communications, Affective Computing
Abstract: The diagnosis of obsessive-compulsive disorder (OCD) often relies on clinical interviews and other subjective assessments. Functional near-infrared spectroscopy (fNIRS) is a non-invasive optical technique widely used to monitor brain activity and assist clinicians in objective brain function evaluation. The advancement of time-series-to-image transformation techniques has opened new pathways for integrating fNIRS with deep learning; however, these methods often suffer from the loss of hemodynamic response function (HRF) information. To address this challenge, we propose a framework for converting fNIRS time-series data into virtual images while preserving HRF activation information. This framework utilizes globally min-max normalized Gramian Angular Fields (GAF) for virtual image transformation, which are subsequently used in deep learning classification models. In cross-validation experiments, the GAF virtual images incorporating HRF information achieved the highest average classification accuracy of 86.7%. Our findings indicate that GAF virtual images integrated with HRF features provide a novel approach for fNIRS-based OCD classification research, further promoting the clinical application of fNIRS technology.

Keywords: Brain-based Information Communications, Cognitive Computing
Abstract: This paper investigates the potential of a minimalist spiking neural network for digit recognition tasks, using the MNIST dataset as a benchmark. The proposed model features a single spiking neuron utilizing the Izhikevich neuronal model, deliberately crafted without weights or a learning phase, embodying the minimalist approach of maximizing performance with minimal resources. Our approach integrates self-organizing maps with a novel optimization method for cluster selection, leveraging Kolmogorov complexity and prototype abstraction. The model achieves 90% accuracy on inverted grayscale Arabic numerals. On Roman numerals, it maintains strong 87.4% accuracy, excelling particularly on well-separated patterns. These results confirms the model's suitability for regular inputs and highlights the potential of minimalist spiking architectures for future neuromorphic systems.

Keywords: Brain-based Information Communications, Medical Informatics
Abstract: Monitoring the Depth of Anesthesia (DoA) using Electroencephalography (EEG) is essential for patient safety and optimal drug administration. However, existing methods face challenges like computational inefficiency and limited generalization, hindering real-time applicability. To address these, we propose a lightweight framework based on SimGate, a parameter-free RNN gating mechanism. SimGate simplifies gating by dynamically computing hidden states based on cosine similarity and requiring only a single linear layer and an initialization vector for training, thus reducing model complexity and enabling efficient parallel computation. To improve generalization, we introduce Harmonic Mix, a frequency augmentation strategy that enhances data diversity by applying harmonic-constrained low-pass filtering and convex combinations of EEG activities. This method preserves critical frequency bands and mitigates noise interference. Experimental results on the VitalDB dataset show that our model achieves 82.70% classification accuracy (ACC) and 5.79±0.64 root mean squared error (RMSE), outperforming existing models with only 16.9K parameters and an inference speed (IS) of 0.02ms. These results demonstrate the effectiveness of our model for real-time DoA monitoring in clinical settings.

Keywords: Brain-Computer Interfaces, Brain-based Information Communications, Design Methods
Abstract: We propose auditory tagging, a novel method to enhance decoding performance in auditory brain-computer interface (BCI) paradigms. Drawing inspiration from steady-state visually evoked potentials (SSVEPs), auditory taggers involve embedding a steady frequency onto an auditory stimulus with the goal of eliciting a detectable neuronal response. In this work, we introduce three such approaches and evaluate them on the auditory intention decoding (AID) paradigm. In AID, subjects are primed with a question and potential target and non-target answer options are provided for this question. The BCI then decodes whether a given sample is a target or non-target. Despite the conceptual promise of the auditory taggers, our experiment results did not reveal statistically significant improvements in decoding accuracy using the proposed tagging approaches. We discuss potential explanations for this observation and highlight possible avenues of improvement for future research.

Keywords: Brain-Computer Interfaces, Brain-based Information Communications, Human-Machine Interface
Abstract: Electroencephalography (EEG)-based wearable brain-computer interfaces (BCIs) face challenges due to low signal-to-noise ratio (SNR) and non-stationary neural activity. We introduce in this manuscript a mathematically rigorous framework that combines data-driven noise interval evaluation with advanced SNR visualization to address these limitations. Analysis of the publicly available Eye-BCI multimodal dataset demonstrates the method's ability to recover canonical P300 characteristics across frequency bands (delta: 0.5-4 Hz, theta: 4-7.5 Hz, broadband: 1-15 Hz), with precise spatiotemporal localization of both P3a (frontocentral) and P3b (parietal) subcomponents. To the best of our knowledge, this is the first study to systematically assess the impact of noise interval selection on EEG signal quality. Cross-session correlations for four different choices of noise intervals spanning from early to late pre-stimulus phases also indicate that alertness and task engagement states modulate noise interval sensitivity, suggesting broader applications for adaptive BCI systems. While validated in healthy participants, our results represent a first step towards providing clinicians with an interpretable tool for detecting neurophysiological abnormalities and provides quantifiable metrics for system optimization.

Keywords: Evolutionary Computation, Hybrid Models of Neural Networks, Fuzzy Systems, and Evolutionary Computing, Deep Learning
Abstract: Distributed deep learning faces a major bottleneck in communication efficiency due to the frequent exchange of gradient information among computing nodes. Gradient compression offers a viable solution to reduce communication overhead. However, existing methods often rely on uniform or limited adaptive strategies, which fail to fully exploit compression potential. To address this, we propose a layer-wise adaptive gradient compression method based on genetic algorithms. By dynamically adjusting compression parameters per layer using evolutionary search and incorporating multi-objective optimization, our approach achieves a better trade-off between compression rate and model accuracy. Experiments across various network architectures and datasets demonstrate that our method significantly improves communication efficiency without sacrificing accuracy.

Keywords: Evolutionary Computation, Metaheuristic Algorithms
Abstract: Dynamic Optimization Problems (DOPs) pose significant challenges because of the evolving nature of their objective functions and constraints over time. These difficulties become more pronounced as the frequency of landscape changes and the dimension of the search space increase. In this work, a novel hybrid approach for dynamic optimization, called Evolutionary Fractal Decomposition based Search (EFDS), is proposed. EFDS uses fractal-based decomposition for space indexing and Evolutionary Algorithms (EAs) to select prominent regions while maintaining population diversity. Experimental results on the Moving Peak Benchmark (MPB) demonstrate the effectiveness of the proposed approach, outperforming competing methods in 21 out of 24 benchmark configurations.

Keywords: Evolutionary Computation, Metaheuristic Algorithms, Computational Intelligence
Abstract: Multi-modal multi-objective optimization has become a hot research topic in the evolutionary multi-objective optimization (EMO) community. To support this line of study, many multi-modal multi-objective test problems have been developed to better understand the behavior of each algorithm in achieving a good balance between convergence and diversity in the decision space. However, the trade-off between decision space diversity and objective space diversity has not been well investigated. In this paper, first, we clearly explain that the widely used DTLZ1–4 test problems exhibit multi-modality (i.e., there exists a many-to-one mapping from the Pareto front to the Pareto set) despite their frequent use as standard benchmark problems for multi-objective optimization. Then, we demonstrate that the trade-off between decision space diversity and objective space diversity can be visually examined using the DTLZ1–4 problems with three or more objectives. Using these four test problems, we examine the search behavior of three standard EMO algorithms and three multi-modal multi-objective evolutionary algorithms (MMEAs). Experimental results show that uniformly distributed solutions obtained by the standard EMO algorithms in the objective space do not have good uniformity in the decision space. In contrast, solution sets obtained by different MMEAs show different trade-offs between decision space diversity and objective space diversity.

Keywords: Evolutionary Computation, Metaheuristic Algorithms, Computational Intelligence
Abstract: NSGA-II, the most well-known evolutionary multi-objective optimization (EMO) algorithm, is widely believed to be ineffective at handling many-objective optimization problems (i.e., problems with four or more conflicting objectives). Since NSGA-II is a Pareto dominance-based EMO algorithm, it loses strong selection pressure as the percentage of non-dominated solutions increases. Additionally, if dominance-resistant solutions (DRSs) exist, they further impair the convergence ability of NSGA-II. Two approaches have been proposed to improve convergence and eliminate DRSs. One is to increase the dominated region of each solution, while the other is to increase the correlation among objectives. These two approaches work well for many-objective optimization and are equivalent under certain conditions. However, they require appropriate parameter values for different problem types. To avoid the difficulty of setting parameter values appropriately, we propose a dual population-based NSGA-II (called DP-MNSGA-II) that utilizes two different sub-populations for objective modification. One subpopulation uses a small parameter value (i.e., small modification), while the other uses a relatively large value for objective modification in NSGA-II. Experimental results show that the proposed DP-MNSGA-II works well on different problems without the need for careful parameter tuning.

Keywords: Evolutionary Computation, Swarm Intelligence, Application of Artificial Intelligence
Abstract: Prompt engineering has become an effective tool for numerous large language models (LLMs)-based tasks. However, it is still a challenge to adaptively optimize the prompt for the target task, where gradients are inaccessible via APIs and the search space is highly complex. Due to the ability to robustly handle complex and multimodal landscapes, and to adaptively explore complex search spaces without explicit gradient information, evolutionary computation (EC) algorithms have garnered significant attention in black-box optimization. However, the performance of different EC algorithms for prompt optimization remains uncertain, which cannot provide guidance for algorithm selection and design in prompt optimization. To alleviate this, a series of investigative experiments are conducted in this paper to evaluate the efficiency of EC algorithms in prompt optimization, which results in three novel findings. First, different EC algorithms, including XNES, CMA-ES, DE, and PSO, are compared on various few-shot learning tasks. The findings indicate that XNES can outperform other algorithms, achieving faster or smoother convergence and higher final accuracy under high-dimensional settings. Second, the role of truncation strategies is also investigated. Results show that moderate bounds effectively help the algorithm balance exploration and exploitation, whereas overly small or large bounds diminish performance. Third, the influence of different k-shot values is examined on the optimization performance for few-shot tasks, which shows that k=32 consistently provides the best trade-off between convergence speed and final model accuracy. In conclusion, these investigations validate the efficacy of the investigated EC algorithms and highlight the advantages of XNES for complex prompt optimization. They further underscore the importance of proper truncation boundaries and k-shot configurations, offering valuable guidance for future black-box prompt tuning scenarios. These insights pave the way for future work on more diverse and complex tasks with larger LLMs.

Keywords: Evolutionary Computation, Swarm Intelligence, Machine Learning
Abstract: In the field of Multi-Agent Reinforcement Learning (MARL), strategy selection and optimization are key challenges in improving agent performance. The Policy Space Response Oracles (PSRO) algorithm is widely used in MARL, and the meta-solver is one of its cores. However, existing meta-solvers may exhibit limitations in complex MARL environments, such as significant consumption of computing resources, bad convergent effect, poor stability, and so on. Therefore, an adaptive multi-layer Prioritized Fictitious Self-Play (PFSP) is proposed in this paper as a meta-solver method for the PSRO algorithm, which further improves the effectiveness of strategy optimization by utilizing the game results of the meta-game in a more efficient and reasonable way. Adaptive multi-layer PFSP can flexibly make strategy choices based on payoff at different layers, thus overcoming the shortcomings of traditional meta-solvers in dealing with complex strategy spaces. The experimental results show that the proposed method significantly improves the convergence speed and performance of strategies in complex MARL environments, especially in the training and testing of the Google Research Football environment, demonstrating its potential and advantages in practical applications.

Keywords: Infrastructure Systems and Services, Fault Monitoring and Diagnosis, System Architecture
Abstract: Serverless computing abstracts infrastructure management, enabling cost-efficient and scalable application deployment. However, multiple operations can be performed simultaneously, leading to system degradation of resources exhaustion during prolonged executions. This study assesses resource utilization in serverless environments using Knative, testing four OS-container engine configurations (Ubuntu/Docker, Ubuntu/Podman, Debian/Docker, Debian/Podman) under stress workload. Results show Ubuntu paired with Docker achieves optimal resource efficiency, avoids software aging, and maintains consistent performance, making it the most effective configuration for scalable serverless deployments. The absence of progressive degradation in resource metrics underscores serverless architectures’ potential to resist software aging through autoscaling. Findings provide actionable insights for optimizing resource management in open-source serverless platforms.

Keywords: Infrastructure Systems and Services, Intelligent Green Production Systems, Technology Assessment
Abstract: In recent years, artificial intelligence (AI) has advanced rapidly, transforming various sectors of society and delivering value to governments, businesses, and end users. However, these advances have also increased computational costs and training times, resulting in significant energy consumption by AI systems. This situation highlights the urgent need for regulatory discussions and proposals that guide AI development and innovation toward energy sustainability. This work reviews, organizes, and categorizes recent regulatory policies for AI energy sustainability into four dimensions for trend analysis. The results show that most identified policies concentrate on transparency and energy efficiency, indicating a higher maturity of these measures within regulatory discussions. This work seeks to raise awareness among stakeholders involved with AI and sustainability, encouraging reflections on energy consumption and efficiency to build a more sustainable future.

Keywords: Infrastructure Systems and Services, Smart Buildings, Smart Cities and Infrastructures, System Architecture
Abstract: This paper introduces EdgeWidgets, a versatile IoT platform designed for environmental monitoring applications. The platform integrates a Bosch BNO055 inclinometer with an ESP32 C3 SuperMini microcontroller and implements a dual-protocol communication layer (HTTP and MQTT). Our experimental evaluation shows that MQTT achieves approximately 26.1% lower latency compared to HTTP, with significantly more consistent performance. These results demonstrate the substantial performance benefits of MQTT’s persistent connection design for monitoring in connectivity-challenged environments.

Keywords: Infrastructure Systems and Services, System Architecture, Large-Scale System of Systems
Abstract: Virtualization platforms like Proxmox VE, which extend hypervisors such as KVM with orchestration tools, are increasingly adopted in high-performance computing (HPC) for improved flexibility and resource management. This study evaluates Proxmox VE against KVM and bare-metal environments to quantify virtualization overheads in CPU-bound and inter-process communication (IPC) workloads. Using texttt{HPL} and texttt{NetPIPE} benchmarks, competitive runs show Proxmox VE’s 0.5%--1.1% compute overhead versus bare-metal (KVM: 0.9%--1.2%), rising to 1.8%--2.0% for cooperative MPI-based texttt{HPL}. Proxmox VE reduces RAM usage by 15%--20% over KVM via dynamic ballooning. Intra-VM IPC achieves native performance (~120~Gbps, <0.005~µs), while intra-host IPC sustains ~50~Gbps (vs. KVM’s 60~Gbps) with 20%--30% latency increases. Inter-host IPC under high contention suffers 30% bandwidth loss and 25%--30% latency degradation. CPU utilization remains stable (~50%) across platforms. Results demonstrate Proxmox VE’s viability for CPU-bound/single-host HPC, with network optimizations needed for distributed workflows.

Keywords: Image Processing and Pattern Recognition, Optimization and Self-Organization Approaches, Biometric Systems and Bioinformatics
Abstract: Multi-Object Tracking (MOT) of polyps in colonoscopy videos serves as a foundational component for intelligent medical workflows. In tracking-by-detection algorithms, managing the equilibrium among competing optimization objectives during data association presents a complex issue. Current approaches, including threshold-based weighted linear aggregation methodologies and hierarchical cascade matching schemes, demonstrate persistent limitations when deployed in endoscopic polyp monitoring scenarios characterized by vigorous polyp movements and complex morphological deformations. Thus, this study proposes a novel polyp tracking framework named MO2Tracker, which can simultaneously optimize multiple objectives during the data association phase. Specifically, MO2Tracker integrates three discriminative similarity cues in parallel: polyp coordinates, detection confidence scores, and appearance consistency, to establish associations between new detections and existing trajectories. To resolve this optimization challenge, we implement Pareto Front analysis via a multi-objective optimization algorithm, deriving Pareto Front solutions. Subsequently, we design two novel selection paradigms, knee-oriented selection and human-inspired selection, to filter optimal detector-trajectory pairings from the Pareto Front. Extensive experiments conducted on two public datasets and one private dataset validate the superior performance of MO2Tracker. On the public dataset SUN-SEG, MO2Tracker achieves improvements of +4.1% in Multi-object Tracking Accuracy (MOTA), +7.2% in Higher Order Tracking Accuracy(HOTA), and +5.8% in IDF1 Score.

Keywords: Biometric Systems and Bioinformatics, Image Processing and Pattern Recognition, Machine Vision
Abstract: Fetal brain automatic segmentation and localization from MRI remains challenging, as existing methods often struggle with multi-view consistency and lack generalization across different fetal types and gestational ages. In this paper, we propose a novel cascade framework Co-UNext for generalized multi-view fetal brain segmentation and identification. Our model is based on a two-step cascade architecture that can take multi-views stacks from the same maternal volume, combining depthwise separable convolutions, attention mechanisms, and other strategies. It achieves precise segmentation of the brain across a set of multi-views stacks and provides fetus-specific segmentation results for multiple pregnancies. We evaluated on 212 fetal MRI scans from 20-36 weeks gestation, including 124 singletons and 88 twins scanned on 1.5T Siemens with axial, coronal, and sagittal views. Results demonstrate that Co-UNext achieves better segmentation performance and can accurately localize all the fetal brains with fetus-specific labels in twin stacks, while also providing more continuous segmentation for data affected by motion and artifacts. On both datasets, the proposed method achieved the best results - 96.9% Dice and 95.3% MIoU on singletons, 96.3% Dice and 94.9% MIoU on twins. Co-UNext shows state-of-the-art generalization capabilities for fetal brain segmentation to facilitate various subsequent quantitative analyses.

Keywords: Computational Life Science, Deep Learning, Transfer Learning
Abstract: Accurate particle tracking is essential for studying intracellular dynamics. While deep learning has advanced tracking performance, its reliance on large labeled datasets hampers generalization to unlabeled biological data. To bridge this gap, we propose a novel particle tracking enhancement framework that designed for annotation-scarce scenarios. At its core is ParticleDiff, a conditional diffusion-based trajectory generator tailored for biological motion. By leveraging historical trajectories as context, ParticleDiff conditionally predicts future positions in an autoregressive loop, ensuring both temporal coherence and biological plausibility. These generated trajectories are used to augment training for downstream tracking models, improving accuracy in data-scarce scenarios. Experiments demonstrate that trajectories generated by ParticleDiff closely resemble real biological data. As a result, tracking models trained on ParticleDiff-augmented data consistently outperform those using synthetic trajectories from other generators, achieving performance comparable to models trained on real annotations. Our framework significantly improves tracking accuracy while reducing reliance on labeled data, thereby enhancing the generalization of deep learning-based tracking models in life science applications. Code is available at https://github.com/imzhangyd/ParticleDiff.

Keywords: Cybernetics for Informatics, Computational Life Science
Abstract: Several prior studies focused on the estimation of the probability of Type 2 Diabetes Mellitus (T2DM) complications. None of the references used sophisticated mathematical models to simulate the long-term progression of the disease when estimating the probability of a complication. This study extends an existing T2DM model to predict the probability of Diabetic Retinopathy. The benefit of this approach is that we will need much less information to predict the probability of a complication than in a standalone model, as the progression of diabetes will be estimated with an identifiable model. To expand the model, a detailed qualitative analysis was required first to write the model equation along with relationships that can be supported by the pathogenesis of diabetes and the biological background of the development of complications. Then the collection of data necessary for setting up the model, which in the course of our present work was implemented by quantitative analysis of publicly available data releases. The next step was to write the model and identify it using the available data. Finally, we validated the model using a data set from a data source independent of the previous ones and evaluated the results. In the course of our present work, we focused on the probability of developing Diabetic Retinopathy. The completed model is suitable for supplementing the simulation of the initial model in such a way as to estimate the probability of retinopathy appearing in a patient during the progression of the disease.

Keywords: System Modeling and Control
Abstract: Handling incomplete data in spatiotemporal systems is a critical challenge in scientific and engineering applications. This paper introduces a time/space twist-based spatiotemporal modeling framework to address incomplete sampling in distributed parameter systems (DPSs). By decomposing system dynamics into temporal basis functions and spatial coefficients, the proposed method reconstructs missing spatial information using neural networks while preserving physical principles. The framework integrates data-driven modeling with physical constraints, ensuring accurate reconstruction even under sparse observations. Experimental validation on a catalytic reaction process highlights the effectiveness and superiority of the approach compared to traditional methods, demonstrating its ability to capture complex spatiotemporal dynamics with limited data.

Keywords: Decision Support Systems, Control of Uncertain Systems, System Modeling and Control
Abstract: The increasing integration of renewable energy sources and battery energy storage systems has amplified uncertainties in power system operations, necessitating advanced estimation methods that transcend traditional quality-oriented approaches. In this paper, we propose a prescription-centric estimation framework that embeds decision-making insights directly into the anticipation process. By leveraging parametric programming and implicit gradient representations, our approach establishes an analytical mapping between uncertain parameters and optimal operational decisions, thereby addressing the inherent asymmetry between uncertainty estimation and system re-balancing costs. Notably, the proposed method breaks through the limitations imposed by linearization constraints in conventional estimation networks and optimization models, paving the way for more accurate and robust decision-making. An iterative bi-level nonlinear optimization strategy is also introduced to overcome the shortcomings of purely data-driven methods. The effectiveness of the framework is demonstrated through case studies, underscoring its potential to enhance both decision accuracy and efficiency in power system operations.

Keywords: Infrastructure Systems and Services, Intelligent Power Grid, Intelligent Green Production Systems
Abstract: This paper presents a new Physics-Informed Neural Network (PINN) framework to estimate the State of Health (SOH) of lithium-ion batteries. The proposed architecture, PINN-LSTM-KF, integrates Long Short-Term Memory (LSTM) networks with extended Kalman filtering under physics-based constraints. Conventional data-driven approaches often fail to generalize across different operating conditions due to nonlinear degradation patterns. Our method addresses these challenges by enforcing electrochemical constraints within a multi-scale architecture. It simultaneously captures microscopic physical processes, mesoscopic temporal dynamics, and macroscopic uncertainty quantification. Experiments on lithium-ion, lithium iron phosphate, and lithium-sulfur batteries demonstrate that the proposed framework achieves Mean Absolute Percentage Errors below 0.01% under physics-informed configurations. Model compression techniques further reduce memory overhead, enabling real-time deployment on embedded systems. The framework also supports feature-level interpretability by quantifying contributions of physical variables to degradation, offering practical insights for battery design and management. These results highlight the potential of combining physics-based modeling with learning-based estimation to improve reliability and safety in energy storage systems, particularly in critical domains such as electric vehicles and smart grids.

Keywords: Electric Vehicles and Electric Vehicle Supply Equipment, Intelligent Power Grid
Abstract: Accurate estimation of the state of health (SOH) of batteries is of great importance for the safe and efficient operation of energy storage systems. However, data-driven methods are often affected by limited generalisation due to sample distribution shift between different batteries, which make them difficult to extent models to unknown domains. To this end, a Transformer-based architecture integrated with RevIN is proposed in this study, which has presented a significant enhancement on the adaptability to distribution shifts of the input data. The normalisation and denormalisation process of Reversible Instance Normalisation reduces statistical bias whilst maintaining trend information. A comparative evaluation is conducted on the NASA battery datasets across six representative models, including Random Forest, eXtreme Gradient Boosting, Multilayer Perceptron, Long Short-Term Memory, Transformer, and the proposed RevIN-Transformer, under both intra-battery and cross-battery prediction settings. Moreover, the analysis of variance is employed to assess the consistency of SOH prediction errors across different batteries. The results indicate that whilst deep learning methods generally outperform traditional models, the RevIN-Transformer method achieves superior accuracy and stability in cross-battery SOH prediction tasks against distribution shifts. In addition, they also validate the effectiveness of integrating lightweight normalisation modules in a Transformer-based architecture for battery SOH time-series forecasting.

Keywords: Smart Buildings, Smart Cities and Infrastructures, Electric Vehicles and Electric Vehicle Supply Equipment
Abstract: One of the most promising flexible sources in decarbonized power systems is based on electric vehicles fleets equipped with V2G technology, which can exchange bidirectional power flows with the grid matching the system operator request, and providing valuable ancillary services, as far as frequency and voltage regulation are concerned. Providing these services requires deploying resilient, highly scalable, plug-and-play, and privacy-preserving computing frameworks for charging/discharging orchestration according to the requested active and reactive power profiles at the point of connections. For this purpose, this paper proposes a decentralized and self-organizing scheme based on a Peer-to-Peer architecture, leveraging privacy-preserving through cooperative consensus protocols. The effectiveness of the proposed scheme is demonstrated by both theoretical and experimental studies, which have been obtained on a hardware testbed simulating a variable number of grid-connected vehicles under realistic operation scenarios.

Keywords: Quantum Machine Learning, Quantum Cybernetics, Agent-Based Modeling
Abstract: Quantum Key Distribution (QKD) leverages quan- tum effects to address the key distribution problem. Tradition- ally, its use has been limited to high-tech specialist networks for dedicated, trusted partners. However, advancements in hybrid quantum-classical networks now suggest the potential for QKD- level security on a wider scale. The Isabelle Infrastructure and Insider Framework (IIIf) has demonstrated the ability to formally verify security properties in proof-of-concept QKD models. In this paper, we build on these foundations to extend the formal analysis to real-world multi-photon QKD implementations. We enhance the existing model and show how IIIf’s attack tree analysis exposes threats such as the intercept-and-resend attack and the photon-number splitting (PNS) attack. Furthermore, refinement in IIIf enables the extraction of a formal specification for decoy-state QKD, strengthening security against vulnerabilities.

Keywords: Machine Learning, Agent-Based Modeling
Abstract: In reinforcement learning (RL), agents maximize accumulated rewards through trial-and-error in the environment to obtain high-performing policies. However, in some situations, loopholes in the purely synthetic reward signals are often exploited by agents, leading to unsafe behaviors, which necessitates the incorporation of safety constraints. In this paper, we propose a safe RL algorithm called Constrained Policy Optimization with Approximately Monotonically Increasing Rewards (CPO-AMIR) to address the safe policy learning in different scenarios and provide practical solutions. We present a novel update formula to achieve a better balance between increasing the reward and decreasing the cost. Furthermore, a theoretical analysis is provided to demonstrate that our approach guarantees the approximately monotonic improvement in rewards when learning constraint-satisfying policies. Our empirical results illustrate the effectiveness and superiority of CPO-AMIR on a set of constrained control tasks.

Keywords: Machine Learning
Abstract: Inspired by diffusion models that revolutionized image generation through language conditioning, offline reinforcement learning (RL) has been reformulated as a sequence modeling problem. Similar to how language descriptions guide image generation, RL sequence models require task-specific conditioning information to generalize to new tasks. We propose Prompt Decision Diffuser (PDD), addressing the generalization challenge in offline meta reinforcement learning (OMRL) by treating it as a sequence modeling problem, with demonstration-based prompting for few-shot adaptation. These prompts, encoded from few-shot demonstrations, effectively capture task-specific information to guide cross-task policy generation. Experimental results on Mujoco and Point-Robot benchmarks demonstrate PDD's superior few-shot generalization capabilities compared to baseline approaches.

Keywords: Deep Learning, Machine Learning
Abstract: Improving lithium battery charging technology is crucial for advancing the widespread adoption of electric vehicles. A well-designed charging strategy should achieve the optimal balance between charging time and battery safety. The learning-based charging method employs an agent to directly interact with the battery plant, iteratively updating the charging strategy based on observations at each time step. This eliminates the need for complex modeling, but strategy performance is highly dependent on the accuracy of observed data. Inevitably, observations of the real battery are often subject to measurement errors and external interference (e.g. adversarial attack), which can potentially lead the agent to make incorrect decisions, posing a threat to the health and safety of the battery. To address this issue, this article proposes a fast-charging method base on robust deep reinforcement learning framework, which considers adversarial perturbations of state observations. The proposed framework is implemented in the deep deterministic policy gradient (DDPG) charging algorithm, wherein the policy network is trained by adding a smooth and robust regularization term. Results show that the proposed method enhances both the smoothness and robustness of the strategy compared with the current DDPG algorithm.

Keywords: Machine Learning, Deep Learning
Abstract: Generative Adversarial Imitation Learning (GAIL) replicates expert behaviors by employing adversarial training involving the discriminator and the generator. In theory, GAIL can balance discriminator and generator through considerable online interactions to learn well. But real-world adversarial training is tricky and unstable. Early mistakes by the discriminator can lead to bad learning outcomes, or the generator might just copy average expert actions (mode collapse) instead of diverse strategies. To address these, we propose Process Adversarial Diffusion Imitation Learning (PADIL). Employing a conditional diffusion model as the generator facilitates the generation of a multi-modal strategy, thereby reducing the likelihood of mode collapse and enhancing imitation learning performance with fewer online interactions. At the same time, it lessens the generator's vulnerability to the fluctuating signals or errors conveyed by the discriminator throughout the training process. Furthermore, we have revised the sample extraction approach of the diffusion policy to tackle the concern of iterative instability under rapidly fluctuating reward signals. Experimental results demonstrate that our method achieves better performance compared to baseline methods.

Keywords: Cybernetics for Informatics, Expert and Knowledge-Based Systems, Information Assurance and Intelligence
Abstract: The proliferation of Internet users has accelerated the spread of fake news on social media, necessitating effective fake news detection systems. However, existing methods often focus solely on the features of claims without considering their uncertainty, limiting their reliability and generalizability. Inspired by Bayesian neural networks, this paper proposes a Bayesian Transformer-based Fake news Detection system with Evidence awareness (BTFDE). To validate the effectiveness of BTFDE, we conducted experiments on several datasets. The results demonstrate that BTFDE outperforms several baseline methods, improving the reliability of fake news detection by quantifying uncertainties and incorporating evidence awareness. This approach improves the generalizability of the model and provides a more rational basis for fake news detection.

Keywords: Cloud, IoT, and Robotics Integration, Cyborgs,, Machine Learning
Abstract: Hand-eye calibration is a crucial step in the implementation of vision-based robotic arm systems. However, existing calibration methods struggle to adapt to scenarios where the robotic arm or vision system frequently changes position. To address this challenge, this paper proposes a novel calibration method based on an exploration-optimized Twin Delayed DDPG (TD3) algorithm enhanced with kernel density estimation and decaying noise, referred to as E-TD5. The proposed approach not only incorporates the E-TD5 algorithm to improve the exploration and exploitation capabilities of the TD3 reinforcement learning framework but also introduces an adaptive target-perception enhancement system to handle frequent variations in hand-eye positioning. Experimental results validate the effectiveness of the proposed method, its robustness to changes in hand-eye positions, and the significant advantages of the E-TD5 algorithm compared to the standard TD3 approach.

Keywords: Information Assurance and Intelligence, Complex Network
Abstract: Cloud computing and edge computing technologies offer effective solutions to the substantial storage and computational demands resulting from the rapid increase in edge network traffic. However, data outsourcing may lead to potential leakage of users' sensitive information due to the limited trustworthiness of cloud and edge systems, allowing malicious users to conspire with these systems to access private data stored in the cloud. To mitigate security risks associated with collusion attacks on existing ciphertext search protocols, we propose a data security search protocol that is resistant to collusion in cloud-edge-end collaboration. This protocol enables fine-grained search of ciphertexts in the cloud while providing protection against collusion attacks. Furthermore, we have conducted formal methods analysis and verification of the protocol, demonstrating its alignment with its intended purpose.

Keywords: Complex Network, Machine Learning, Representation Learning
Abstract: Identifying key nodes in a graph is critical for the analysis and management of networked systems. Traditional centrality-based methods primarily focus on the intrinsic characteristics of nodes and the topological properties of graphs. However, these approaches depend heavily on handcrafted feature selection, resulting in poor generalization across networks with diverse structures. Graph Convolutional Networks (GCNs), as a classical deep learning model for graph-structured data, have demonstrated promising performance in key node identification. By aggregating features from a node and its neighbors, GCNs construct expressive node representations and enhance the model’s generalization ability. Nevertheless, existing GCN-based methods often fail to adequately capture the relative importance of neighboring nodes. To address this limitation, this paper introduces a novel key node identification model based on Graph Attention Networks (GAT), termed KeyGAT. By leveraging an attention mechanism, the proposed model enables adaptive and reliable fusion of node and neighbor features, thereby improving identification accuracy. Experimental results on different graph datasets validate the effectiveness of the proposed approach.

Keywords: Complex Network, Optimization and Self-Organization Approaches, Deep Learning
Abstract: As 6G technology continues to evolve, future communication systems demand extremely low latency, improved energy efficiency, and support for massive device connectivity. To address these demands, the integration of Space-Air-Ground Integrated Networks (SAGIN) with Mobile Edge Computing (MEC) has emerged as a compelling strategy. This integration leverages edge nodes deployed on low Earth orbit (LEOs) satellites, unmanned aerial vehicles (UAVs), and terrestrial small base stations (SBSs) to deliver distributed computational resources. Nevertheless, due to the inherent heterogeneity and limited capabilities of these edge devices, reliably offloading computing tasks from ground user equipments (UEs) to appropriate edge nodes—whether satellite-based, aerial, terrestrial, or via local execution—remains a significant challenge. In this paper, we design a metadata-driven intelligent offload prediction and global resource optimization framework for centipede games, the metadata only contains the key information of the computing task, but not the actual task data itself. Then we envision a 6G smart city network architecture with complex computing scenarios, formulate the problem of minimizing the global average delay and energy consumption as a Markov Decision Process (MDP) by combining with Bellman's optimization equations, and propose an asynchronous metadata driven centipede game method (AMDCG) based on deep reinforcement learning (DRL). Simulation results show that the AMDCG method significantly reduces the system offloading overhead in terms of latency and energy consumption compared to other benchmark algorithms.

Keywords: Complex Network, Swarm Intelligence
Abstract: With the advancement of artificial intelligence and UAV technologies, UAV swarms have been increasingly applied in a wide range of diversified missions. However, existing studies remain limited in their characterization of UAV swarm under dynamic mission demands and multi-layer complex interactions, as well as in the quantitative evaluation of their reliability. To address these gaps, this study proposes a mission-oriented multi-layer super-network modeling and reliability evaluation framework. First, four heterogeneous sub-networks are constructed from the dimensions of operation, mission, communication, and resource, capturing the swarm’s multi-layer interaction relationships. Then, considering inter-layer dependencies and cascading failures, a random failure and network reconfiguration strategy is introduced to quantitatively analyze the impact of key node failures on topological metrics and network vulnerability. Finally, a fire rescue case study is conducted to verify the effectiveness and accuracy of the proposed method.

Keywords: Cognitive Computing, Augmented Cognition
Abstract: The adapter technique has significantly improved the performance of fine-tuning pre-trained language models (PLMs) in multi-task settings, especially for small models with limited resources. However, current adapter frameworks typically rely on single-task data, a fixed representation space, and a single training phase. This approach limits their ability to generalize across domains in multi-task scenarios, thus restricting performance improvements for smaller models. To address these challenges, we propose the cross-domain contrastive optimization (CDCO) framework to enhance performance in multi-task learning. CDCO improves model performance by asynchronously co-optimizing across diverse task data sources, representation spaces, and multi-stage structures. Specifically, CDCO introduces innovations in both data sample selection and training strategy. First, CDCO introduces out-of-domain manifold sampling (ODMS), which enhances training diversity by selecting challenging hard-negative samples from out-ofdomain datasets through manifold learning. Second, CDCO employs multi-stage asynchronous co-optimization (MAC), mapping samples from ODMS to Euclidean and Poincar´e spaces. Then, it constructs a cross-domain contrastive loss based on the spatial properties of these distributions to guide the optimization process. By sequentially optimizing adapter layers across different spatial distributions, CDCO maximizes the potential of the adapter while mitigating overfitting, thus improving the adaptability and stability of small models in multi-task environments. Experimental results demonstrate that CDCO significantly improves performance on in-domain (ID), out-of-domain (OOD), and knowledge-intensive (KI) tasks, confirming its broad applicability and effectiveness.

Keywords: Cognitive Computing, Human Performance Modeling
Abstract: This study investigates the task of genuine-fake word classification in the Chinese language, with the objective of comparing human lexical decision-making with that of fine-tuned transformer-based language models. We propose an integrated framework that combines behavioral experiment data with multiple fine-tuned transformer-based models, enhanced with downstream classification layers. Experimental results show that these models consistently outperform human participants in accuracy, precision, and F1-score across all word frequency levels, with particularly greater stability in recognizing low-frequency words. From a cognitive perspective, the findings support the word frequency effect and information processing theory, indicating that low-frequency items impose higher cognitive load. Human participants may mitigate uncertainty through semantic compensation and risk-sensitive strategies. This study not only confirms the effectiveness of fine-tuned language models in lexical anomaly detection, but also provides empirical insights into the cognitive mechanisms underlying human word recognition.

Keywords: Cognitive Computing, Human-centered Learning, Environmental Sensing,
Abstract: Growing mental health problems among college students underscore the urgent need to enhance the restorative effect of university campus. However, current evaluation tools lack systematic, scalable, and interpretable methods to quantify the psychological impact of open space design. This study proposes a multimodal perception system to predict the restorative effect of university open spaces by integrating visual, structural, and semantic features. We construct a large-scale dataset comprising 600 campus images and 12,147 subjective ratings collected through standardized psychological scales. Semantic segmentation is used to extract spatial visual indices from images, while semantic impressions are obtained via the Semantic Differential (SD) scale and converted into natural language descriptions. A dual-encoder alignment framework maps both index and text representations into a shared latent space, enabling cross-modal prediction of restorative scores. A Random Forest regressor is trained on this space to support score inference from either image- or text-based inputs. In addition, we apply a Rule-based Representation Learner (RRL) to extract interpretable spatial patterns associated with restorative outcomes. Experiments show that our method significantly outperforms traditional regression models, achieving an R^2 of 0.85 in predicting perceived restorative effect. The learned rules reveal both explicit visual drivers (e.g., greenery, sky openness) and implicit spatial logics (e.g., element interaction). This framework offers a lightweight and interpretable evaluation tool for health-oriented campus design, applicable across design and planning stages even without image data.

Keywords: Cognitive Computing, Human-Machine Interaction, Interactive and Digital Media
Abstract: Chinese characters represent a unique ideographic writing system where glyph structures encode rich semantics, yet current NLP models struggle to leverage these form-meaning associations. This paper introduces Radical-Graphormer, a novel computational framework addressing this gap by modeling the systemic relationship between character component structure and semantics. Using heterogeneous graph representations within a variational autoencoder (VAE) backbone, the framework uniquely captures the morpho-semantic symbiosis at the radical level. It enables bidirectional tasks: interpreting semantics from glyph structure and generating structurally-valid glyphs from semantic inputs. Supported by HanziFormSem, a new large-scale form-meaning dataset of 3,587 characters (details provided), our model effectively learns structure-to-meaning mapping (Macro-F1 0.81 for semantic classification). Crucially, it achieves controllable, semantics-driven glyph synthesis, generating characters with high structural validity (95%) and visual credibility (4.1/5.0 human rating). This work demonstrates the feasibility of computationally simulating Chinese character composition, offering a new methodology for modeling complex symbolic systems and opening avenues in computational linguistics, AI-driven design, and digital humanities.

Keywords: Human Perception in Multimedia, Intelligence Interaction, Cognitive Computing
Abstract: 大型语言模型在艺术描述方面表现出令人印象深刻的流畅性，但区分大型语言模型生成的文本和人类文本之间的语义抽象和创造力仍然是一个挑战。本研究通过多维度分析，包括词向量、句法、情感、隐喻检测和跨模态对齐，系统地比较了人类和大语言模型生成的书法描述。提出了一种优化的基于BERT的分类器和BLIP2闭环实验，以捕捉风格和语义上的区别。结果表明，人工智能生成的文本表现出更高的句法复杂性（平均句子长度：18.2 vs. 13.0）和词汇多样性（10.07 vs. 5.26），但在创造力（组内相似度：0.781 vs. 0.505）和跨模态任务的语义减少（周期准确率：21.7% vs. 35.3%）方面滞后。虽然人工智能在结构化表达和一致性方面表现出色，但它仍然受到训练模式的限制。相比

Keywords: Human Performance Modeling, Cognitive Computing, Human-Computer Interaction
Abstract: Survey research is a vital tool for understanding human opinions, behaviors, and preferences, but traditional methods face challenges such as high operational costs, limited sample diversity, and privacy concerns. Large Language Model (LLM)-based agents offer a novel approach to simulate human responses, generating high-quality synthetic data that preserves privacy while enabling researchers to explore patterns in public opinions. Our study introduces a flexible LLM-based survey simulation platform, which personalizes responses using real-world social media data to enhance realism and relevance. We conducted a validation case study using data from Steemit users, comparing the personalized simulated responses generated by LLM agents with actual survey responses collected from these users. This comparative analysis revealed key differences and potential biases, providing insights for refining survey methodologies. Personalized LLM simulations were highly effective for factual questions but faced limitations with subjective topics, often demonstrating lower variance compared to human responses. The quality of social media data and careful model selection also significantly influenced the accuracy of simulations. Our platform enables social science researchers to explore new methodologies for survey design, pre-test the impact of framing, and interact dynamically with LLM agents at reduced cost. This work provides valuable insights into the use of LLM-based agents for enhancing survey research, supporting their application in social science.

Keywords: BMI Emerging Applications, Passive BMIs, Active BMIs
Abstract: In recent years, the number of traffic accidents caused by the errors of older adults while driving in Japan has been on the increase. As we age, changes in reaction time can make it more challenging to drive safely. We aim to develop a system that estimates the driver’s intention and assists the driver in driving accordingly. We investigated whether the detection of movement-related cortical potential, which is a component in electroencephalography (EEG) that occurs before driving, would enable us to detect a driver’s intention to steer. In this study, we conducted experiments to measure EEG signals while a participant drove on a simulator. We analyzed the results using machine learning and classified the EEG during straight driving and before and after steering. Because the optimal learning model for a classifier to detect Movement-Related Cortical Potential from EEG has not yet been derived, we trained Light GBM, LDA, SVM, and LSTM models and identified an appropriate classifier. The result reveals that the best accuracy was obtained with Light GBM, which was able to detect steering intentions with an accuracy of approximately 60%. Further improvements in estimation accuracy and the development of a driver assistance system based on steering intention are future tasks.

Keywords: Other Neurotechnology and Brain-Related Topics, BMI Emerging Applications, Passive BMIs
Abstract: Fatigue poses a significant challenge to safety and performance across various domains, making its accurate detection and classification essential for preventing accidents and enhancing operational efficiency. Fatigue classification using an electroencephalogram (EEG) plays a critical role in applications including occupational health, transportation safety, and performance monitoring in high-stakes environments. Deep learning models, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), have shown potential for analysing EEG data but often face limitations in capturing both temporal dependencies and long-range patterns critical for fatigue detection. This study investigates the use of a transformer-based temporal convolutional network (TCN-Transformer) for EEG-based fatigue classification, leveraging the TCN's causal and dilated convolutional architecture in conjunction with transformer components to efficiently extract multi-scale temporal features and model long-range dependencies in sequential data. Comparative experiments demonstrate that the transformer-enhanced TCN framework outperforms CNN and RNN models for real-time fatigue classification, achieving an accuracy of 87.44%, a sensitivity of 93.27%, and an AUC of 0.9680. The evaluation strategy, employing 10-fold cross-validation repeated over 10 runs, validated the robustness of the proposed approach. This study establishes transformer-based TCNs as a promising tool for EEG-based fatigue classification, providing a streamlined and effective method for real-time fatigue monitoring in safety-critical environments.

Keywords: BMI Emerging Applications, Other Neurotechnology and Brain-Related Topics
Abstract: Covert visuospatial attention (CVSA) enables users to direct focus without eye movements, offering a promising non-invasive control modality for brain-machine interface (BMI) systems. Traditional CVSA-based BMIs rely on decoding α-band electroencephalography (EEG) activity from whole-trial data, often overlooking the temporal dynamics of attentional processing. In this study, we propose a novel control framework that explicitly separates the CVSA task into two subcomponents: shift attention and sustained attention. EEG signals were recorded from nine participants performing a CVSA task in a real-world environment. Through spectral feature analysis and temporal segmentation, we identify distinct patterns associated with each subcomponent and train dedicated classifiers accordingly. Our pseudo-online analysis suggests that combining the outputs of two task-specific classifiers significantly improves both trial-level classification accuracy (by an average of +23.74%) and area under the curve (AUC) (by +4.49%) compared to the conventional approach. These findings suggest that modelling the temporal structure of CVSA can enhance decoding performance and offer a more robust foundation for attention-based BMI.

Keywords: Passive BMIs, BMI Emerging Applications, Other Neurotechnology and Brain-Related Topics
Abstract: Attention is a critical cognitive function in daily life, and its impairment can lead to serious consequences. Electroencephalogram (EEG) enables detection of brain activity related to attention, but decoding attentional states from EEG signals remains challenging due to noise artefacts and low spatial resolution. While deep learning approaches have shown promise, their generalizability across subjects remain limited. This work proposes a novel end-to-end hybrid deep learning architecture, STESA-Net, for classifying attention versus inattention states from multichannel EEG signal. STESA-Net integrates spatiotemporal convolutional layers, self-attention mechanism, and Bidirectional Long Short-Term Memory (Bi-LSTM) units to effectively capture and weight relevant spatial and temporal features. The design enhances the model’s ability to learn subject-specific patterns while maintaining generalizability across individuals. The proposed method is validated using 11 subjects EEG data recorded during a simulated driving task, employing leave-one-out-validation (LOOV) method. The proposed method achieved an average attention v/s inattention classification accuracy, precision, recall and F1 score of 78.47%, 80.16%, 78.46%, and 78.14%, respectively. The improvement in average classification accuracy achieved by the proposed method over the existing methods is ~7%. The results highlight the effectiveness of combining convolutional layers, self-attention mechanisms, and Bi-LSTM networks for robust cross-subject attentional state classification. The proposed framework offers a promising solution for real-world applications in attention monitoring and may aid the early detection of attention-related cognitive impairments with reduced computational overhead and enhanced generalization.

Keywords: BMI Emerging Applications, Passive BMIs
Abstract: EEG-based identification of brain states has advanced significantly, enabling cognitive monitoring during daily tasks. Brain dynamics can be studied as a Markovian, Semi-Markovian, or Non-Markovian stochastic process, prescribed by Transition Probability Matrices derived from EEG measurements.

This study models brain dynamics as discrete Markov chains using second-by-second dominant frequencies derived from the power spectrum of high-density array EEG signals. By analyzing transition and limiting probabilities across modalities, we reveal distinct neural signatures that differentiate engaged from meditative states. Though preliminary, these findings highlight the method’s potential to enhance BCI systems and deepen our understanding of brain dynamics, brain health and the benefits of meditation in future studies.

Keywords: Active BMIs
Abstract: Covert attention to peripherally presented visual stimuli is an effective cognitive task that can be used to operate a brain-computer interface (BCI) even if the user is completely paralyzed. Brain responses to single visual search display onsets show attention-related components emerging between 180ms and 500ms. This poses a challenge when using fast visual stimuli that induce steady-state visual evoked potentials (SSVEP) for covert attention BCIs since a new stimulus is presented before the current stimulus is processed. Here we investigated brain activity during covert attention by presenting visual stimuli inducing SSVEPs in the left and right visual field. We decoded the attention shift to in-phase stimuli in the left and right visual field with an average accuracy of 72.5% across all channels. When left and right stimuli were presented with a phase shift but constant frequency, the average decoding accuracy increased to 84.6%. While with in-phase stimuli, top-down attention is decoded, decoding of attention to antiphase stimuli additionally takes advantage of stimulus-related processing.

Keywords: Visual Analytics/Communication, Team Performance and Training Systems, Human Factors
Abstract: Meetings are important activities that influence the achievement of team objectives. The characteristics of creative meetings must be clarified to enhance creativity. This study conducted two analyses with the aim of identifying participant behaviors that contribute to creativity by quantitatively analyzing creativity in actual meeting data. First, utterances in meetings were annotated using 11 categories, and frequently occurring dialogue patterns were examined both overall and for each team through pattern analysis. Second, regression analysis was conducted to examine the relationship between the frequency of these frequent patterns and a quantitative creativity index constructed based on text sentiment. In the first analysis, the most frequently observed type of interaction in meetings was the exchange of opinions. Additionally, communication characteristics specific to each team were identified. However, no significant relationships were found between the creativity index and the dialogue patterns in the second analysis. Further analysis, including a reconsideration of the method of aggregating conversation patterns, is required in future work.

Keywords: Medical Informatics, Affective Computing
Abstract: “Multi-modal AI-Based Pain Detection in Inter- mediate Care Patients in the Postoperative Phase” is an interdisciplinary research work that operates in the domain of automated pain detection. It aims to improve previous work, based on pain databases like BioVid and UNBC shoulder pain, as well as AI-based approaches using computer vision and signal processing to analyze available modalities. Thus, we present our basic research idea on how to improve automatic pain detection in three major steps. The first step focuses on collecting pain data from postoperative patients in intermediate care stations (IMC). In addition, patients who are not fully oriented should be included in a separate data collection as a second focus group. Then, improvements on the state-of-the-art models should not only advance general pain detection, but also help bridge the gap to the real-world setting of the IMC data. Improvements include transferability analysis, feature selection evaluation, and balancing of data distribution to deliver better classification performance. In a last step, we aim to test, verify and evaluate the classification performance on the IMC data with the support of medical practitioners.

Keywords: Human Factors, Human-Machine Interaction
Abstract: Self-efficacy is crucial for the effective application of assistive technology and rehabilitation. This study proposes a novel approach to assess the impact of motor interventions on motor self-efficacy, relevant for human-robot interaction in rehabilitation, by focusing on the perceived reachable space. Twelve healthy adults underwent an arm movement restriction intervention using a robotic arm (KINARM), and changes in the perceived reachable space and muscle activity were measured before and after the intervention. The results indicated a reduction in the perceived reachable space and an adaptive decrease in muscle activity for unreachable targets following motor restriction. This suggests that the perceived reachable space can serve as an objective proxy for task-specific motor self-efficacy, which is valuable for evaluating user adaptation to robotic interfaces. Furthermore, these findings imply that in rehabilitation using interactive robots, a patient's effort levels may be influenced by their perception of task achievability.

Keywords: Human-Machine Cooperation and Systems, Human-centered Learning, Kansei (sense/emotion) Engineering
Abstract: In the automotive industry, improving vehicle operability tailored to individual drivers is essential for achieving a long and enjoyable driving experience. This study presents a demonstration of a personal-fit control system that adaptively adjusts vehicle characteristics to match each driver’s preferred operability. The experiment is conducted using a driving simulator. In addition, an evaluation method based on a Kalman filter is examined. The results indicate that optimal vehicle characteristics vary among drivers, and that the personal-fit control system enables adaptation to these individual preferences.

Keywords: Human-centered Learning, Cognitive Computing, Human Factors
Abstract: In a rapidly evolving global context - driven by technological innovation and societal change - the need for continuous reskilling and upskilling in education and training is more urgent than ever. To address this challenge, CONSALE (Constructing Situation Awareness in microLearning Environments) offers a structured framework for adaptive microlearning that aligns instructional goals with cognitive processes. By integrating Understanding by Design with Situation Awareness-Oriented Design, CONSALE enables the creation of personalized, context-aware learning experiences. Its implementation in Moodle — via a plugin-based architecture — supports dynamic learner profiling (based on the Felder-Silverman model), behavioral adaptation, and cognitively tagged content delivery, enhancing engagement and learning outcomes.

Keywords: Virtual/Augmented/Mixed Reality, Systems Safety and Security, Human Factors
Abstract: With the proliferation of mixed-reality (MR) systems in aerospace and defense, there is increased potential for adversarial exploitation of system vulnerabilities and/or properties in the human cognitive process in order to reduce mission-effectiveness. This paper presents our preliminary work on the Modeling and Analysis Toolkit for Realizable Intrinsic Cognitive Security (MATRICS), a formal methods-based approach to provide a mathematically rigorous design and verification framework for protecting MR systems and operators in mission-critical applications from cognitive attacks. We describe our approach and present initial results, including formal models of the human operator, MR device, and mission environment, and apply existing formal methods tools to prove the holistic cognitive security of MR systems.

Keywords: Human-Machine Interaction, Assistive Technology
Abstract: Accurate decoding of human motion intention from surface electromyography (sEMG) signals recorded non-invasively from the skin surface is critical for enabling intuitive control in assistive robotics and human–machine interactions. With the advancement of high-density sEMG (HD-sEMG), neural decoding methods based on motor unit (MU) activity have shown promise due to their potential to capture finely controlled movement information. However, the effects of data segmentation parameters on the decomposition and decoding accuracy remain underexplored. In this study, we systematically investigated how the segmentation length and data size of sEMG signals used for decomposition affect the performance of finger force decoding. Specifically, HD-sEMG signals were recorded from eight human participants during single- and multi-finger isometric force tasks. A neural decoding pipeline was developed for finger force predictions. We first evaluated the impact of four segmentation window lengths (10 s, 20 s, 40 s, and 80 s) on decoding accuracy, and found that a 20-second window was sufficient to ensure accurate decoding, with no additional benefit from using longer segments. Using this setting, we further examined the effect of training data size by comparing decoders trained with different data sizes. Our results showed that using the full training dataset significantly improved decoding performance compared to using only half of the training dataset. These findings offer practical guidelines for optimizing data usage in MU-based motion intention decoding systems.

Keywords: Human-Machine Interface, Supervisory Control, User Interface Design
Abstract: With the ongoing digital transformation of industry, web-based human-machine interfaces (HMIs) are gaining increasing importance. In motion control systems, HMIs play a vital role in facilitating user interaction for monitoring and control. Recent advancements in web technologies have spurred interest in adopting modern web-based interfaces for motion control systems, as they not only enhance user experience but also facilitate the integration of advanced technologies such as machine learning. This paper presents a case study on the design and implementation of a web-based HMI for CNC (computer numerical control) and proposes machine learning-based security enhancements. The system is developed using Python and Flask for the backend service, and Node.js with Vue.js for the frontend client interface. LinuxCNC, an open-source motion control software, serves as the control engine, while communication is established via WebSocket and RESTful APIs to exchange command and status. Experimental evaluation using industrial motor drives and controllers demonstrates that the proposed system offers comparable performance. Compared with the legacy native HMI, while the web-based HMI exhibits slightly higher input latency due to network overhead, it significantly reduces output latency while improving responsiveness. To address security concerns, the paper introduces an AI-driven framework that integrates LSTM (long short-term memory) based anomaly detection and interaction biometrics for implicit user authentication. A customized design is proposed to ensure secure operation within web-based motion control environments.

Keywords: Human-Computer Interaction, Interactive and Digital Media
Abstract: This paper investigates the interpretive processes within interactive art, focusing on post-participation interviews from an immersive installation featuring a mirror that displayed a previous participant instead of the viewer. The study, conducted through semi-structured interviews, explores how participants interpret and construct personal meaning from their engagement with the artwork. Thematic analysis revealed key areas of interest present thorough the participants’ responses. Additionally, participants found metaphorical discussions more insightful than titles in framing meaning. The findings are analyzed through a framework for the psychology of aesthetics, offering new insights into participant experience in interactive art.

Keywords: Intelligence Interaction, Human-Computer Interaction, Ethics of AI and Pervasive Systems
Abstract: Large Language Models (LLMs) are widely used today in virtual assistants and content generation. However, there are suspicions that LLMs present confirmation bias, responding in a way that reinforces beliefs or assumptions embedded in users' questions, which can lead to erroneous decision-making, especially in sensitive areas such as healthcare. The objective of this research is to determine how often and under what conditions LLMs present confirmation bias and to identify the causes of this effect. The methodology involves conducting an experiment in which 52 biased healthcare questions are presented to 10 of the most popular models and analyzing whether their responses were biased. This work proves with statistical power the behavior of confirmation bias. We show that confirmation bias in LLMs occurs in all LLMs with a frequency of 20% to 60% of the occasions. The evidence suggests that the bias arises from the training database, the Transformer architecture itself, and the instructions in the fine-tuning phase by the companies behind the LLMs. This research explores pathways for the development of trustworthy LLMs.

Keywords: Human-Machine Cooperation and Systems
Abstract: Despite recent advancements, Large Language Models (LLMs) face difficulties in providing reliable and verifiable information, especially in critical domains. It is challenging to find possible hallucinations or manage the intrinsic knowledge of modern models. This paper presents an approach leveraging human-machine cooperation inspired by LLM technologies and Knowledge Graph (KG) structures. The proposed framework enables users to create a hierarchical network of concepts from textual input. Concepts range from characters (low-level) through words (medium-level) to word chains (high-level). The system generates relationships mainly on word co-occurrences within sentences, while leveraging Natural Language Processing (NLP) techniques and statistics for providing recommendations for entities and relations that users then validate and curate through an interactive interface. This cooperation between human and machine embraces the strengths of computational accuracy and storage capability with human cognitive skills. The idea aims to empower users to effectively organise complex information and navigate the information space with ease through the help of strongly connected concepts.

Keywords: Information Visualization, Human-Machine Cooperation and Systems, Cognitive Computing
Abstract: Accurate 3D semantic labeling is critical for scene understanding and navigation in autonomous driving systems. However, manual annotation of 3D point clouds is labor-intensive and difficult to scale across diverse environments. To address this challenge, we propose a modular-based automated labeling framework that leverages 2D semantic cues to facilitate 3D scene understanding. The system integrates state-of-the-art models—Grounding DINO, SAM2, YOLOv7, and Generative AI (GAI) —to perform high-quality 2D semantic segmentation, which is then processing into semantic LiDAR point clouds. Afterward, the framework employs multi-frame fusion and static-dynamic scene separation to construct dense 3D semantic occupancy grids. Additionally, GAI serves as a prompt-refinement module, improving segmentation accuracy at object boundaries and in distant regions. Experiments on real-world street-view datasets demonstrate that AutoLabel-2D significantly enhances labeling efficiency and segmentation completeness, offering strong generalization across scenes. This framework provides a scalable and effective solution for high-definition mapping and semantic perception in autonomous driving applications.

Keywords: Human Performance Modeling, Human Factors, Human-Centered Transportation
Abstract: Evacuation behavior is an important human factor in safety performance modeling, and pre-evacuation phase is an interval between receiving an alarm signal and decisive escaping for safety. To describe human awareness and response during pre-evacuation phase, this paper formulates a multi-agent process to simulate crowd pre-evacuation dynamics. The model mainly combines classical opinion dynamics with binary phase transition to describe how group pre-evacuation time emerges from individual interaction in a given social context. The model parameters are quantitatively meaningful to human factors research within socio-psychological background, e.g., to what extent an individual is stubborn or open-minded, or what kind of the social topology exists among the individuals and how it functions in aggregating individuals into groups. The modeling framework also describes collective motion of many agents in a planar space. The resulting multi-agent system is similar to Vicsek flocking model, and it is meaningful to explore complex behavior during phase transition of non-equilibrium processes.

Keywords: Human Perception in Multimedia, Brain-Computer Interfaces, Brain-based Information Communications
Abstract: Understanding how the brain processes musical pitch errors is key to advancing music perception and auditory rehabilitation. This pilot study (N=1) examined neural responses to pitch deviations in a trained musician using a portable dry-electrode electroencephalography (EEG) system. The participant listened passively to seven~10-second violin excerpts with systematically varied pitch errors: none, medium, or high. EEG data were recorded, preprocessed, and analyzed for oscillatory patterns in the beta band across temporal and central brain regions. Findings revealed a graded increase in right-temporal beta power with error magnitude, alongside distinct temporal dynamics suggesting a sequential process: initial detection of deviations, evaluation of error severity, and adjustment of predictive models. These results highlight a potential cascade in passive pitch-error processing, extending prior work on active performance to naturalistic listening contexts. This work offers preliminary insights into the neural basis of pitch perception and suggests applications in music education and auditory training, though further research with larger cohorts is needed to confirm these findings.

Keywords: Assistive Technology
Abstract: With the advancement of hearing aids and cochlear implants, auditory perception among the Deaf and Hard-of-Hearing (D/HoH) people has improved, including the recognition of environmental and instrumental sounds. However, even with these improvements, it remains difficult for many deaf individuals to perceive emergency vehicle sirens. This study proposes more perceivable siren patterns by modifying pitch and tempo. Through awareness surveys and listening experiments with D/HoH individuals, we examined the perceived urgency and audibility of modified sirens. The results suggest that increasing the tempo improves both recognizability and perceived urgency, while higher octave shifts tend to be rated as noisy or annoying.

Keywords: Assistive Technology, Human Enhancements, Design Methods
Abstract: This paper addresses the design of an assistance control system that enables safe and flexible operation while facilitating operator upskilling. A key factor in upskilling is increasing opportunities for operators to attempt control actions. To this end, we design an assistance system that reduces intervention frequency and promotes operator freedom. We particularly focus on designing a human assistance system for water level control in a tank system. Then, we characterize the set of admissible control actions for the human operator in order to design assistance control logic that ensures the safety of the overall system. Finally, we conduct a human-in-the-loop simulation of the tank system. Through the simulation, we verify that the proposed assistance system ensures the safety of the overall system with minimal intervention, demonstrating its potential as a foundational technology for effective operator upskilling.

Keywords: BMI Emerging Applications, Passive BMIs, Other Neurotechnology and Brain-Related Topics
Abstract: Prolonged use of brain–computer interfaces (BCIs) with virtual reality (VR) via head-mounted displays (HMDs) induces mental fatigue, potentially impairing neurorehabilitation. This study examines EEG-based fatigue markers in healthy participants during extended BCI-HMD sessions. Fatigue was classified using N-way Partial Least Squares (N-PLS) with linear discriminant analysis, achieving 82.42% (±7.5) accuracy. N-PLS components revealed spatial–spectral patterns in occipital and sensorimotor alpha activity. Temporal trajectories indicated progressive fatigue accumulation during sessions. Results demonstrate the feasibility of EEG-based fatigue monitoring for optimizing BCI-HMD post-stroke neurorehabilitation.

Keywords: Robotic Systems, Manufacturing Automation and Systems
Abstract: Industrial robots are typically designed with a fixed, standardized structure, which often forces manufacturers to use a system that's over-dimensioned, or far larger and more powerful than needed for the task at hand. This not only results in unnecessary costs but also takes up valuable floor space. A new approach, known as modular robotics, offers complete freedom in robot configuration. By using scalable, easily pluggable motor and link modules instead of a rigid, one-size-fits-all design, machine builders can create customized robot solutions with the precise kinematics required for a specific job. Combined with an open, PC-based control technology that provides programming options for both beginners and experienced PLC programmers, the ATRO system from Beckhoff Automation is intended to be a foundational step toward this future of modular industrial robotics.