Nonlinear constrained optimization is really at the heart of computational mathematics, ranging from resource allocation to material stress analysis. The contribution of this paper is a sophisticated optimization framework that harnesses trust-region methods, dynamic regularization, and thorough sensitivity analysis to solve high-dimensional nonlinear problems. The proposed method minimizes an objective function comprising quadratic terms, linear penalties, and sparsityinducing regularization, subject to stress and budget constraints. The framework achieved precise convergence in 46 iterations and 756 function evaluations, satisfying the gtol termination condition with optimality of 4.43 × 10−9 and zero constraint violations, all within 0.36 seconds. To provide insight into the optimization process, the following four visualizations are generated: a Convergence Plot showing exponential convergence of the objective function that converges and stabilizes after approximately 100 iterations; a Gradient Magnitude Plot showing exponential decay in the gradient norm, confirming smooth and efficient convergence; a Hessian Matrix Heatmap showing the structure of the dependencies and curvature around the optimal solution to assist in sensitivity interpretation; and a Variable Contribution Bar Plot, which quantifies the relative contribution from each decision variable to the objective function to identify the dominant contributors. This work showcases the robustness, efficiency, and scalability of the proposed framework. The visualizations not only validate the results of the optimization but further give insight into the solution space. These results open up perspectives toward the extension of such a framework to more complex dynamic constraints and challenging real-world datasets, bringing important contributions to computational mathematics and applied optimization.

This paper presents a detailed examination of Group Key Establishment methods, pivotal in securing group communication over public networks. We classify Group Key Establishment into three key categories: classical, quantum-safe, and quantum-based, each tailored to address distinct challenges in key management, particularly under the constraints of member dynamics and resource efficiency. Highlighting the critical influence of emerging quantum technologies, our study emphasizes the growing importance of quantum-resistant approaches to safeguard group communications against quantum threats. This paper traverses the evolution of Group Key Establishment protocols, shedding light on the shifting cryptographic paradigms and future prospects in the era of quantum advancement.

Rapidly identifying anomalous events on freeways can significantly improve emergency response time and reduce congestion. However, the unpredictable nature of these incidents, along with the limitations of current monitoring systems, makes accurate and timely detection challenging. In this study, we propose a Cycle-Consistent Bidirectional Graph Generative Adversarial Network (CCB-GraphGAN) model to enhance lanelevel traffic anomaly detection on freeways. By treating freeway traffic information as graph-structured data, our model captures the joint distribution of regular traffic patterns and their underlying latent representations through an adversarial process. It also incorporates cycle consistency constraints to effectively reconstruct both normal traffic data and its latent features during training. During the inference phase, we employ an autoencoder-based anomaly detection approach to identify lanelevel traffic anomalies. We evaluate the performance of our method through comparative and case study analyses using the recently released Freeway Traffic Anomalous Event Detection (FT-AED) dataset. The results indicate that our method outperforms all baseline models in detecting freeway anomalous events early. Notably, it achieves a lower percentage of missed crashes compared to baseline models. With a false positive rate (FPR) of only 1%, our method reduces detection delay by identifying crashes an average of 5 minutes earlier than their official report time. Additionally, an analysis of real-world crash events further confirms our method's effectiveness in accurately identifying freeway anomalies at an early stage.

Neural Token Compression (NTC) introduces a dynamic tokenization strategy that adjusts token representations based on semantic and structural dependencies, aiming to enhance the efficiency and scalability of Large Language Models (LLMs). Traditional tokenization methods often result in fixed token sets, which may not optimally capture the complexities of diverse linguistic patterns. NTC addresses this limitation by enabling adaptive token merging, thereby reducing the number of tokens processed during training and inference. This reduction leads to decreased computational overhead and memory usage, facilitating the deployment of LLMs in resource-constrained environments. The implementation of NTC involves a mathematical framework that leverages semantic similarity measures and structural dependencies to inform token merging decisions. Experimental evaluations demonstrate that NTC achieves significant improvements in perplexity and compression ratios compared to baseline tokenization approaches. Qualitative analyses further reveal that NTC maintains the consistency and coherence of generated text, indicating its potential for various downstream tasks. However, certain limitations, such as variability in token sequences and the complexity of capturing linguistic dependencies, warrant further research. Future directions include refining the NTC framework and exploring its applicability across diverse languages and domains. The integration of NTC into LLM architectures signifies a meaningful advancement in natural language processing, offering a promising avenue for optimizing tokenization processes and improving the performance of largescale language models.

One of the main goals of wireless sensor networks is to permit the involved nodes to communicate with low energy budgets, as they are typically battery-powered. When such networks are employed in industrial scenarios, constraints about latency may have a significant role, too. The TSCH mechanism, and more in general TDMA schemes, rely on traffic scheduling, and consequently they can feature low power consumption and more predictable latency. Some recent proposals like PRIL-M enable further consistent energy savings, but unfortunately they cause at the same time a dramatic increase in latency. This work presents an extension of PRIL-M, we named PRIL-ML, that achieves a significantly shorter latency in exchange for a slight increase in power consumption. Its operating principles are first illustrated, then some approximate equations are provided for assessing analytically the improvements it achieves, starting from simulation results obtained for both standard TSCH and the original PRIL-M technique.

Abnormal traffic patterns caused by extreme events have the potential to disrupt traffic flow on large regions of urban road networks. Timely and reliable detection of such disruptions is crucial for effective traffic management. However, existing methods for detecting extreme traffic events are unable to simultaneously identify disruptions at both network and local levels. Moreover, most existing methods cannot handle incomplete traffic data, which can be an essential challenge during extreme events. In this study, we address these limitations by creating a novel method for traffic anomaly detection using a multitask sequence-to-sequence learning process. Specifically, we propose a Self-Imputing Deep Multitask Sequence (SI-DMSeq) model that can simultaneously reconstruct sequential traffic data and provide one-step-ahead predictions. This model also integrates an unsupervised data imputation strategy to enable training with incomplete data. During inference, the proposed model adopts an autoencoder-based anomaly detection approach to identify large-scale disruptions at both network and local levels. Additionally, it utilizes one-step-ahead predictions to reliably impute missing data under typical traffic conditions. To evaluate the performance of our proposed method, we applied it to the Manhattan road network in New York City using historical traffic data. The results indicated that the proposed method accurately detects disruptions caused by Hurricane Sandy in 2012 at both network and local levels and efficiently imputes missing traffic data under typical conditions.

Recently, Photo-Voltaic energy has been a subject of intensive research for green and sustainable energies. The boost converter is the main backbone of this conversion. Conventionally, coupled inductors have been employed to increase and extend the output voltage of this converter. This work introduces a novel converter using a conventional boost combined with a coupled inductor. The proposed topology responds to a wider range of overcasts and functions as an efficient harvester with an improved output current in Continuous Current Mode (CCM). Furthermore, it extends the functionality of the harvester by improving the efficiency and output voltage concurrently. A prototype was built and successfully tested to verify the proposed topology. An efficiency of 91% is retained during a partial overcast under matched condition, with a single stage converter.

This study introduces a novel multi-sensor system and a fusion strategy for continuous ice thickness detection. To address the limitations of traditional labor-intensive methods, which have lower resolution and are affected by environmental dependencies, we developed an enhanced air-coupled Ground Penetrating Radar (GPR) system integrated with temperature sensors and a transformer-based algorithm for anomaly compensation. It offers a reliable method for mapping ice structures by leveraging the dielectric contrast between ice and water. However, real-time monitoring during rapid temperature fluctuations remains challenging. The proposed method leverages Dynamic Time Warping (DTW) to identify significant signal changes, while the transformer network offers a physical information-supported prediction for ice thickness using environmental parameters. Results demonstrate that the proposed method improves detection accuracy, providing early warning capabilities and enhancing ice thickness tracking in dynamic conditions. The system real-time performance is validated through extensive field data from the Yellow River, showcasing the effectiveness of this integrative framework in challenging monitoring scenarios.

The field of wireless communications has traditionally been defined by what seems like unending exponential traffic growth. History suggests this trend is unlikely to continue in perpetuity, at least with the current set of applications, with recent evidence pointing to moderating traffic growth. In this article we evaluate the implications of the peak smartphone era for those designing wireless networks, within the context of the next-generation of wireless broadband technologies. Firstly, three potential future demand scenarios are identified ranging from a return to exponential traffic growth (optimistic), through to continued moderation in growth (realistic), and even a scenario of declining traffic (pessimistic). Secondly, we compare the emerging properties of the 6th generation of cellular technology ("6G") envisioned by IMT2030, and two new Wi-Fi standards, including IEEE 802.11be ("Wi-Fi 7") and IEEE 802.11bn ("Wi-Fi 8"). Finally, an alternative vision for the future of wireless broadband is proposed focusing on enhanced coverage, reduced deployment costs, and improved energy efficiency. Four key recommendations include (i) neutral hosts for superior indoor coverage, (ii) ensuring spectrum sharing and intelligent handover/roaming integration between cellular, Wi-Fi and Non-Terrestrial Networks (NTNs), (iii) strong support for infrastructure sharing and national roaming in rural and remote areas, and (iv) efficient (re)organization of existing spectrum allocations.

The rapid development of Electric Vertical Take Off and Landing (eVTOL) aircraft is transforming urban aviation, introducing new challenges for airworthiness certification an essential process to ensure operational safety and regulatory compliance. This paper presents a Type Certification framework for eVTOL aircraft, leveraging the Arcadia Model-Based Systems Engineering (MBSE) approach through the Capella tool. This framework replaces traditional document-based methods with a model-driven approach that integrates regulatory standards and Means of Compliance (MoCs) into a unified system model. Additionally, automation is introduced through Large Language Models (LLMs) to dynamically predict and validate MoCs from textual requirements, streamlining checklist generation and enabling rapid adaptation to evolving regulations. While the framework improves efficiency, it complements rather than replaces expert supervision, emphasizing the importance of human oversight in complex regulatory contexts. The proposed framework is scalable and adaptable, applicable not only to various eVTOL platforms but also to other emerging aircraft types. Findings from case studies and expert evaluations demonstrate a significant reduction in workload, enabling faster certification while maintaining accuracy and regulatory alignment. This study offers a replicable, efficient model for certification that fosters the adoption of emerging technologies while ensuring high safety standards and supporting sustainable growth in urban air mobility.

Neural adaptive deep brain stimulation (NaDBS) using beta band power as a biomarker is currently being investigated as a therapy for Parkinson's disease (PD). These systems modulate DBS amplitude in response to beta power fluctuations resulting from medication use or the underlying disease. While NaDBS has been shown to be a potentially effective treatment for PD, it may not be well suited to capture and respond to the stochastic symptoms of the disease. External kinematic sensors can reliably capture these stochastic symptoms and be used as an additional input to make stimulation decisions. We demonstrate a distributed, hybrid controller that can analyze kinematic inputs from external sensors and neural inputs to make stimulation decisions on a second-by-second basis. The hybrid controller simultaneously streams kinematic data from wearable sensors on the shanks of a participant and neural data from an implantable neurostimulator into a PC in the loop. These data streams are time synchronized on a second-bysecond basis and the two data streams are analyzed. When a freezing event is detected in the kinematic data, the controller rapidly increases stimulation to a predetermined maximum intensity (Imax). When no freezing event is detected, or if the controller is uncertain if a freezing event occurred, the controller modulates stimulation slowly based on beta band power received from the neural input. A benchtop demonstration of this system shows that this controller architecture is feasible. The mean overall system latency was 311ms (SD = 102ms) and the mean latency across the PC in the loop portion was 178ms (SD = 86ms), below the second-by-second basis that the controller operates on. This proof-of-concept hybrid controller can be used in the future to develop DBS control policies with multiple additional inputs.

In the data-driven era, graph data is widely distributed in multiple fields, and its complex network relationships pose higher challenges to data analysis models. Although graph neural networks (GNNs) perform well in processing non-Euclidean spatial data, traditional methods are usually limited to capturing local information through message passing or relying on graph diffusion to obtain global information. This single perspective makes it difficult to achieve a balance between global and local information, especially in heterogeneous graphs, which can easily lead to information loss or over-smoothing. To this end, we propose Constrained Diffusion Graph Neural Networks (CDnet), which optimizes the graph diffusion range by adding dual constraints: Constrained Diffusion (CD) and Feature Information Flow Routing (FIFR). The CD module reduces the diffusion depth by adaptively reducing the number of diffusion steps, efficiently captures global information and reduces invalid propagation. The FIFR mechanism dynamically regulates the direction and strength of information flow between nodes to ensure precise control of information transmission and retention of local features. Extensive experimental results on public homogeneous and heterogeneous datasets show that CDnet outperforms state-of-the-art methods, demonstrating its excellent generalization and robustness in complex graph analysis. Code is publicly available at: https://github.com/FangjingLi/CDnet.

This paper investigates near-field spot beamfocusing (SBF) using holographic antennas for multiuser communication applications. While far-field holographic surfaces have been widely explored, a comprehensive study of nearfield holographic surfaces, especially for precise beam-focusing, remains limited. In this work, we derive a closed-form analytical solution for the focal gap, enabling precise beam-focusing at the intended focal point. We characterize key physical parameters of near-field holographic antennas, including the depth of focus (DoF), angular beamwidth, and the effects of surface impedance quantization. Additionally, we introduce an interference cancellation algorithm that leverages orthogonal beam patterns on the focal plane, enhancing secure communication in multiuser scenarios. Extensive EM simulations and experimental measurements validate the proposed methods, demonstrating effective beam-focusing and interference mitigation for singleand multiuser applications. The results suggest that holographic SBF could be a pivotal technology for next-generation 6G wireless networks, offering significant potential for applications in secure communication, wireless power transfer, and medical technologies.

The era of Quantum is approaching faster than expected due to recent advancements in the field. As it comes closer to reality, data encryption becomes a major issue. The need for quantum-resistant cryptography algorithms has increased. The National Institute of Standards and Technology (NIST) under the United States Department of Commerce has issued a set of standards. Algorithms that pass each of these stages are approved and are called Post Quantum Cryptography (PQC) Algorithms [1]. Currently at the fourth round of its standardization process, NIST has approved four algorithms. These algorithms have been widely implemented on software platforms, but their hardware acceleration still needs a lot of work. Latest developments in the field of FPGAs aids seamless implementation and promise a hardwaresoftware co-design. This article explores the implementation of NIST-approved PQC algorithms on the AMD PYNQ-Z2 Board using Python [2]. PYNQ Z2 is AMD's open-source project that provides us a platform to integrate programs in Python with an FPGA. Stress tests have been performed on this design and its metrics have been stated. [3] The Future scope of PQC algorithms in an FPGA has been summarized.

3D object detection plays a crucial role in autonomous driving, with Bird’s Eye View (BEV) becoming increasingly popular for its rich contextual information, ease of multimodal fusion, and scalability. Despite its advantages, current BEV-based 3D detection methods still face significant challenges, including multi-modal fusion, communication bottlenecks, robustness under varying conditions, and safety concerns. This survey provides a systematic review of recent advancements in BEV perception, organizing the research based on these focal areas. It spans a broad range of perspectives, offering valuable insights for future perception research. Additionally, this survey explores the influence of emerging technologies, such as large language models and end-to-end frameworks on enhancing BEV perception capabilities, focusing on improving performance and robustness. Key future directions include: (1) advancing from isolated vehicle perception to V2X cooperative perception; (2) evolving from single-modal to integrated multi-modal fusion; (3) shifting from simulated environments to real-world applications; and (4) transitioning from hierarchical perception frameworks to interpretable, end-to-end large-scale models.

The swift integration of emerging technologies such as artificial intelligence (AI), quantum computing, and blockchain into workplace environments presents unprecedented security challenges that transcend traditional cybersecurity paradigms. This paper introduces CIPHER (Cognitive Integration Process for Harmonising Emerging Risks), a novel cognitive mental model designed to assist security professionals in navigating the complex and dynamic security landscape of technology-driven workplace change. CIPHER differentiates itself from current frameworks by providing a flexible, cognitive approach to security strategy that can adapt to the unpredictable dynamics of integrating various technologies into companies. The approach consists of six stages to the mental model: Contextualise, Identify, Prioritise, Harmonise, Evaluate, and Refine. CIPHER integrates principles from cognitive science, game theory, and dynamical systems theory to offer a memorable and flexible conceptual framework for assessing and addressing security threats in high-uncertainty, low-information contexts inside linked technology ecosystems. This research illustrates CIPHER's ability to connect theoretical security principles with actual execution via its cognitive foundations, integration with organisational procedures, and applicability across diverse developing technological sectors. The paper examines essential elements of developing technology security, encompassing the ethical ramifications of AI algorithms, privacy and legal issues, and the wider social effects of employment automation. This research demonstrates how the CIPHER mental model can assist organisations in formulating comprehensive, adaptive, and ethically sound security strategies for the swiftly changing environment of workplace technology through theoretical foundations, practical applications, and hypothetical case studies.

Microsoft's large language model (LLM) ecosystem, experienced as Copilot in this paper, increasingly dominates professional communication across government and enterprise sectors. Australia now faces a crisis that current analytical frameworks struggle to fully capture or articulate. Through rigorous analysis of empirical evidence from the Australian Public Service, including a comprehensive trial of Microsoft 365 Copilot (n=5,765) and union survey data (n=1,778), this paper identifies an emerging form of corporate colonisation that operates simultaneously across technological, cognitive, and epistemological domains. The empirical evidence, drawn strictly from these official sources, reveals that 81% of knowledge workers are developing unofficial adaptation strategies to corporate-built and designed AI systems, while 77% express concerns about erosion of public trust through AI use. The data indicates a trajectory towards reduction in linguistic variation across departments, suggesting systematic standardisation of professional communication patterns. More fundamentally, Australia's response to this cognitive colonisation, exemplified by its eight AI Ethics Principles, reveals a dangerous deterioration in the nation's capacity for sophisticated technological analysis. This paper argues that without immediate recognition of the full scope of this crisis, Australia risks not just ceding control of its professional linguistic future to corporate entities but losing the very capacity to understand or articulate the nature of this loss.

Network Digital Twins (NDTs) function as virtual replicas of real networks, enabling real-time monitoring and analysis of 5G core networks. Graph Neural Networks (GNNs) have emerged as a promising approach for node failure classification within NDTs. However, the issue of graph imbalance which adversely affects GNN performance, particularly for minority classes has received limited attention in existing research. Information propagation among nodes can amplify majority class dominance, resulting in reduced classification accuracy. To address this challenge, we propose a Graph Fourier Transform with a Cost-Sensitive Message-Passing Neural Network (GFT-COSMEP) for imbalanced node classification. This approach first applies the Graph Fourier Transform (GFT) to generate node embeddings, followed by message aggregation on a sampled graph to enrich node representations. A cost-sensitive learner then leverages these embeddings to improve imbalanced learning performance. Sensitivity analysis confirms that tuning the model's cost-sensitive parameters enhances accuracy, achieving optimal results at a specific balance of loss terms. Experiments on real-world 5G network and NDT-based failure classification datasets demonstrate that GFT-COSMEP outperforms existing methods, achieving up to a 21.1% accuracy improvement at a 0.3 imbalance rate. These results highlight GFT-COSMEP's effectiveness in managing class imbalances in graph-based data.