Current trajectory datasets of traffic participants often lack detailed environmental information, which is crucial for developing effective data-driven methods for future mobility solutions. To address this gap, we introduce the comprehensive DLR Urban Traffic dataset version 1.2.0. The dataset includes 32,296 trajectories of traffic participants, along with traffic light data, local weather data, air quality data, and road condition data collected at the Application Platform for Intelligent Mobility Research Intersection during a single day of recording. A comparison with other publicly available datasets reveals that our dataset offers more comprehensive information about the traffic environment than existing alternatives. An analysis of our dataset shows that trajectories are available for all possible 16 routes at the intersection, with the number of trajectories per route varying significantly, ranging from 11 to 3,344. Most interactions between motorized road users occur during unprotected left turns with oncoming traffic. However, there are also interactions between motorized and vulnerable road users, particularly during right turns. All in all, the dataset provides researchers with the resources needed to improve urban mobility solutions.

Precise and efficient control of permanent magnet synchronous motors (PMSMs) is critical for many industrial and transportation applications, particularly in electric vehicle drives. Recent advances in data-driven control using deep reinforcement learning (RL) techniques have enabled automated control designs within minutes without requiring plant-specific a priori knowledge. However, these RL techniques typically lack generalizability across motor instances with differing parameters, necessitating retraining for each new system. Meta-reinforcement learning (MRL) extends the capabilities of standard RL by enabling controllers to adapt to new tasks, i.e. different motors, with little to no retraining. This work introduces a universal control algorithm based on deep q-learning direct torque control (DQDTC), extended with MRL techniques, termed meta-DQDTC. Applied to a diverse set of differently parameterized PMSMs, and equipped with online system identification capability, meta-DQDTC performs on par with individually trained RL controllers that are specialized for a single drive. This effectively synthesizes a universal PMSM controller that nearly instantly adapts to different motors without additional training. A key advantage is the significant reduction in controller synthesis time: traditional model-based methods might require days of manual effort, whereas trained only once the meta-DQDTC can be deployed within seconds. Experimental results demonstrate that meta-DQDTC can be successfully transferred to a realworld system, despite being trained exclusively on simulated motors with idealized assumptions. Notably, the specific real-world experiment has herein not been simulated during training, validating the generalization and transfer capability of the meta-DQDTC.

True time delay circuits are generally realized using variable delay all-pass filters due to their flat delay-frequency response. However, they are power hungry and practically, variations in transistor characteristics and parasitic impedances result in delay variations with frequency. This paper discusses a first-order RC-low pass filter as a TTD circuit, which leverages the fact that the phase response starts changing linearly 1 decade before the pole location. The variable delay is accomplished by changing the circuit's effective impedance using the Miller effect obviating the need for varactors or variable resistors. A bandwidth extension of the delay cell is achieved using a delay correction circuit. A 4-bit TTD circuit realized by cascading 15 stages of variable delay LPFs has been designed in TSMC 65 nm CMOS technology and post-layout simulation results show that the TTD has a power consumption of 36 mW and a delay range of 400 ps with a delay variation of ±5 ps across a bandwidth of 1 GHz.

Post-mortem iris recognition brings both hope to the forensic community (a short-term but accurate and fast means of verifying identity) as well as concerns to society (its potential illicit use in post-mortem impersonation). These hopes and concerns have grown along with the volume of research in post-mortem iris recognition. Barriers to further progress in post-mortem iris recognition include the difficult nature of data collection, and the resulting small number of approaches designed specifically for comparing iris images of deceased subjects. This paper makes several unique contributions to mitigate these barriers. First, we have collected and we offer a new dataset of NIR (compliant with ISO/IEC 19794-6 where possible) and visible-light iris images collected after demise from 259 subjects, with the largest PMI (post-mortem interval) being 1,674 hours. For one subject, the data has been collected before and after death, the first such case ever published. Second, the collected dataset was combined with publicly-available post-mortem samples to assess the current state of the art in automatic forensic iris recognition with five iris recognition methods and data originating from 338 deceased subjects. These experiments include analyses of how selected demographic factors influence recognition performance. Thirdly, this study implements a model for detecting postmortem iris images, which can be considered as presentation attacks. Finally, we offer an open-source forensic tool integrating three post-mortem iris recognition methods with explainability elements added to make the comparison process more human-interpretable.

Developing and validating new AI concepts can be prohibitively expensive when relying solely on real-world data collection. This paper proposes a flexible simulation-to-real framework that leverages cost-effective game-based environments as an initial testbed for rapid AI hypothesis testing. Our approach employs a modular ensemble of specialized Convolutional Neural Networks for robust feature extraction, while another simple yet efficient classifier (e.g. Random Forest classifier) provides high-level recommendations. Through iterative evaluation and refinement in simulation, only the most promising methods advance to resource-intensive real-world trials. By reducing costs and accelerating development, this framework is particularly well suited for diverse applications ranging from assistive technology to robotics. Empirical results demonstrate strong accuracy across multiple tasks and underscore the effectiveness of a modular design that can be easily adapted to new domains. The proposed pipeline ultimately aims to streamline AI research by enabling faster decisions about which innovations merit real-world deployment. The code is available at the following repository: https://github.com/sqdArtemy/mspv

The correct design of leakage inductances for multi-winding transformers can be challenging due to the fact that all the windings are interacting. This parameter plays a major role in the operation in isolated power converters such as dual active bridge or multiple active bridge power converters. The existing analytical and numerical methods of estimation of the mentioned leakage inductances are limited in accuracy and do not allow to simultaneously design the device according to the required parameters, as well as to predict the value of leakage inductance. In this paper, a comprehensive methodology has been developed for the design of a multi-winding transformer with accurate leakage inductance determination during the design process. Within the methodology, several stages are proposed: analytical, numerical, validation and prototyping. Each of the stages allows to accurately determine the leakage inductance and to achieve the best possible fabrication quality of the prototype device.

The emergence of large language models (LLM) has driven a significant increase of AI workload in data center power demand. Renewable-powered solutions to decarbonizing LLM workload and reducing electricity costs are faced with the combined volatility of stochastic user requests and renewable energy. The key to removing the barriers in sustainable AI development lies in the adjustable capability of LLM power profiles. Therefore, this paper focuses on exploring the potential load flexibility of LLM workload and proposes a coordinated performance-safe scheduling framework. Driven by the existence of the most energy-efficient GPU core frequency that is smaller than the default maximum one, the framework slows down the fine-tuning cluster and utilizes idle GPU resources from the inference cluster to maintain the computing performance of fine-tuning tasks. Consequently, the power consumption of the total cluster is reduced, which provides a fresh source of load flexibility. Furthermore, the framework employs dynamic frequency scaling to more flexibly modify the power profile of the expanded fine-tuning cluster. The computing performance is particularly guaranteed through temporal coupling constraints. In a simulated study supported by real-world data, the results prove a 6% power-saving ability and 9% cost-saving gains at utmost.

Reflector antennas often exhibit frequency ripples in key performance metrics, including aperture efficiency and antenna noise temperature. These ripples stem from corresponding fluctuations in the electric field pattern, due to interference between fields radiated, reflected or diffracted by the feed and reflectors, and it is important that they are characterised. In this paper, we introduce an innovative approach for reconstructing the electric fields of reflector antennas using simulated data across a range of frequencies. By leveraging the power of sub-Nyquist sampling, the number of required samples can be reduced, leading to increased computational efficiency. This presents an advantage to traditional methods such as the Fourier transform that requires Nyquist sampling. We demonstrate how these reconstructions can be used to characterise the frequency contents of resulting reflector properties, such as aperture efficiency and antenna noise temperature.

Objective: Investigating consistent methods and metrics for classifying Detrusor Overactivity (DO) events and developing an automated robust method for clinical measurements calculation from cystometry data in persons with spinal cord injury (SCI). Methods and procedures: A two-stage method for was proposed to detect DO events. In the first stage, DO peaks were detected using local linear models combined with thresholding criteria derived from clinical definitions and known artifacts. In the second stage, a segmentation method was proposed to detect the start and end time points of each DO event, marking the DO activity periods. As a result, complete clinical measurements, including bladder compliance, can be estimated automatically. The method was developed and tested on 77 anonymized urodynamic samples from SCI patients (40 DO-positive, 37 DO-negative) with 158 annotated DO events. Results: On test data, in terms of the patient-level diagnosis of DO, the proposed method achieved a sensitivity and specificity of 100%. Individual DO event detection achieved an average precision of 0.94 and recall of 0.72. Detrusor activity period identification showed a precision of 0.86 and a recall of 0.88. The task of automated bladder compliance estimation showed that the point-value-based method yields a lower median absolute error compared to the proposed line-fitting-based method, with a median absolute error of 5.20 and 7.14 ml/cmH2O, respectively. Finally, for classifying bladder function into normal, low and severely low compliance, the proposed method had an accuracy of 88%. Conclusion: Our proposed local model fitting with thresholding based on clinical knowledge, achieved accurate automated results for cytometry data, which will enable objective assessment of routinely performed examinations. Clinical and Translational Impact Statement: This work proposes a fully automated detrusor overactivity diagnosis and feature extraction method. Empowering medical teams to consistently assess urodynamic studies while aiding disease characterization and enhancing clinical decision-making for SCI patients. Furthermore, it provides a mathematically defined method for extending the pipeline to other populations and standardizing clinical assessments. Category: Clinical Engineering, Medical Devices and Systems.

The ability to provide reliable data rates across several coverage areas establishes massive multiple-input multipleoutput (m-MIMO) cell-free (CF) systems as a pivotal technology for future sixth-generation (6G) systems. CF networks do, however, introduce additional security and network integrity vulnerabilities. For that, to complement the traditional cryptographic algorithms, physical layer security (PLS) can be an effective strategy for acquiring essential wireless network information in order to develop authentication methods against impersonation attacks. To prevent these spoofing attacks, we propose leveraging the wireless channel and the access point selection (APS) allocation schemes as authentication mechanisms. Our approach begins with a threshold-based analysis of spectral efficiency (SE) losses across different APS schemes. We then propose an algorithm capable of estimating the number of eavesdroppers executing active attacks while identifying the targeted user equipment (UE). Finally, we evaluate the robustness of our detection scheme by evaluating both SE losses and the achievable secrecy SE of a single attacker's position relative to the targeted UE. Results demonstrate that monitoring these parameters provides critical insights into network performance and the impact of active eavesdropping. These findings highlight the potential of integrating PLS with upper-layer authentication protocols to significantly enhance wireless network security.

Co-pilots are AI chatbot agents powered by LLMs, designed to assist users with complex tasks through interactive, natural language guidance. As sensor data becomes more prevalent, its limited insights can be enhanced by LLMs' reasoning capabilities. We envision sensor data co-pilots that transform sparse data into meaningful, interactive descriptions, helping users make informed decisions. By analyzing contributions from devices within defined proxemic zones-considering spatial relationships between users and devices-these co-pilots serve as intuitive interfaces for various applications. For example, in smart homes, a sensor data co-pilot could clarify energy consumption, suggest optimizations, or warn of unusual activity. This user-centered approach helps individuals navigate and comprehend their environments seamlessly. We predict sensor data co-pilots will be essential for existing and emerging technologies, enhancing user interactions and increasing adoption and retention of pervasive devices. As they evolve, these co-pilots will unify sensor data and AI, advancing the vision of transparent, pervasive computing.

This research delves into the optimization of payment gateway systems in the fintech industry using AI-augmented OCR and intelligent workflow automation, especially with a focus on Azure Functions. The paper explores how these tools can make payment processing more efficient, scalable, and cost-effective by combining serverless computing with AI-driven technologies. Cases from financial institutions adopting Azure services, which indicate KPIs for transaction speed, operational efficiency, and total cost of ownership. Surveys from IT experts and system architects give real-world insight into how AI and OCR apply in the payments system.The event-driven architecture of Azure Functions features AI-powered Optical Character Recognition (OCR) for the automatic processing of payment information and intelligent workflows optimizing decision making. It would improve processing time, eliminate human errors, and make transactions' data accuracy more efficient. In addition, features such as Azure Active Directory, encryption, and compliance with PCI DSS standards ensure the safe handling of sensitive payment data, thereby enhancing the security of these systems.The paper compares AI-augmented OCR and intelligent workflows with traditional server-based models with respect to metrics involving transaction speed, system scalability, and cost-effectiveness. Preliminary findings suggest that AI-and OCR-based integration with Azure Functions reduces costs, enhances scalability, and strengthens security in payment gateway systems. These technologies bring about a revolution in payment processing in fintech and improve the efficiency and save operational expenses.

Data analytics has become a cornerstone of effective decision-making in the modern business landscape. Management consulting firms are using data-driven insights to guide organizations through complex challenges and foster innovation. This review paper explores the integration of data analytics within management consulting, analyzing its transformative impact on decision-making processes. The paper discusses methodologies, tools, and best practices by synthesizing the findings from academic studies, industry reports, and practical case studies that consulting firms have employed in this regard. Emerging trends and opportunities are highlighted in the study to provide an overall roadmap for data analytics to help organizations drive strategic initiatives and operational efficiencies. The abstract underlines the importance of data analytics in reshaping traditional consulting practices. It exhibits its potential in optimizing decision-making, enhancing organizational strategies, and providing evidence-based solutions to modern challenges. It thus previews the multifaceted applications and transformative impacts discussed in the paper.

An essential task of channel modeling for vehicular communications and sensing applications is predicting scattering from traffic participants, here represented by complex-shaped curved bodies. One approach is to discretize the curved body into facets and then use Ray Tracing (RT), combined with the Uniform Theory of Diffraction, to find the scattered field. We show that more comprehensive diffraction methods are required to achieve realistic results: one of them is vertex diffraction, and its implementation in the RT framework is considered in this work. Simulations by thus augmented RT are compared with simulations by MLFMM and PO solvers for discretized spheres and a simplified car, respectively.

This paper presents a lightweight modified rod antenna with tapered feeding, designed for high-power microwave (HPM) applications. The proposed antenna is capable of transmitting narrowband high-power pulses centered at 2.45 GHz. Numerical simulations were conducted to assess the antenna's performance, followed by prototype fabrication and experimental validation. The manufactured antenna achieved a bandwidth of 3.07 GHz (3.52 GHz-6.59 GHz) and a total gain of 5 dBi. The absence of dielectric materials and the designed 50ohm input impedance facilitate direct connection to an HPM generator, making the proposed design highly suitable for defense and radar applications.

The forthcoming wireless network is expected to support a wide range of applications, from supporting autonomous vehicles to massive Internet of Things (IoT) deployments. However, the coexistence of diverse applications under a unified framework presents several challenges, including seamless resource allocation, latency management, and systemwide optimization. Considering these requirements, this paper introduces WIND (Wireless Intelligent Network Digital Twin), a self-adaptive, self-regulating, and self-monitoring framework that integrates federated learning (FL) and multi-layer digital twins to optimize wireless networks. Unlike traditional digital twin (DT) models, the proposed framework extends beyond network modeling, incorporating both communication infrastructure and application-layer DTs to create a unified, intelligent, and contextaware wireless ecosystem. Besides, WIND utilizes local machine learning (ML) models at the edge node to handle low-latency resource allocation. At the same time, a global FL framework ensures long-term network optimization without centralized data collection. This hierarchical approach enables dynamic adaptation to traffic conditions, providing improved efficiency, security, and scalability. Moreover, the proposed framework is validated through a case study on federated reinforcement learning for radio resource management. Furthermore, the paper emphasizes the essential aspects, including the associated challenges, standardization efforts, and future directions opening the research in this domain.

This study presents a novel partial differential equation (PDE)-based model for neural stem cell (NSC) therapy in spinal cord injury (SCI) treatment. Existing models often fail to accurately represent NSC migration, differentiation, and interaction with biomaterials and immune responses. By integrating scaffoldmediated diffusion and immune response dynamics, the proposed model offers a more realistic framework for regenerative therapy optimization. The model improves upon previous approaches by providing precise control over growth factor distribution, ensuring sustained bioavailability for enhanced NSC survival and functional integration. Applications of this framework extend beyond SCI treatment, with potential implications for neurodegenerative diseases, brain injury repair, and personalized medicine. Future research will focus on experimental validation and machine learning-driven optimization to refine the predictive capabilities of the model.

Many people use online platforms such as Instagram and Facebook on a daily basis. These applications often feature comment sections where users can express their opinions. Unfortunately, it is common for this feature to be used inappropriately, with people posting offensive comments towards individuals or groups. Defining what represents offensive content is difficult, as it varies by context, making AI removal challenging. In order to offer additional awareness on hateful and abusive conduct in social media, this work focuses on the SemEval-2023 task of binary sexism detection. Our data analysis indicates that adding a new label would be beneficial for enhancing the dataset's alignment with real-world applications. Towards this end, this work presents a simple baseline model using cosine similarity to assess comment alignment with different platform guidelines to highlight how these comments differ from the community guidelines.