Pure OAI Repository
Not a member yet
421387 research outputs found
Sort by
NeRF-CA:Dynamic Reconstruction of X-ray Coronary Angiography with Extremely Sparse-views
No full textDynamic three-dimensional (4D) reconstruction from two-dimensional X-ray coronary angiography (CA) remains a significant clinical problem. Existing CA reconstruction methods often require extensive user interaction or large training datasets. Recently, Neural Radiance Field (NeRF) has successfully reconstructed high-fidelity scenes in natural and medical contexts without these requirements. However, challenges such as sparse-views, intra-scan motion, and complex vessel morphology hinder its direct application to CA data. We introduce NeRF-CA, a first step toward a fully automatic 4D CA reconstruction that achieves reconstructions from sparse coronary angiograms. To the best of our knowledge, we are the first to address the challenges of sparse-views and cardiac motion by decoupling the scene into the moving coronary artery and the static background, effectively translating the problem of motion into a strength. NeRF-CA serves as a first stepping stone for solving the 4D CA reconstruction problem, achieving adequate 4D reconstructions from as few as four angiograms, as required by clinical practice, while significantly outperforming state-of-the-art sparse-view X-ray NeRF. We validate our approach quantitatively and qualitatively using representative 4D phantom datasets and ablation studies. To accelerate research in this domain, we made our codebase public: https://github.com/kirstenmaas/NeRF-CA.</p
The sound of progress:AI voice agents in service
No full textPurposeThis article advances research on artificial intelligence (AI)-powered voice agents (hereafter, AI voice agents) by providing (1) a conceptualization of AI voice, (2) a conceptual mapping of AI voice agents, and (3) an organizing framework outlining key benefits, risks, and contingency factors associated with AI voice agents in service. This introduction to the special issue on “Voice Capabilities in Smart Service Systems” also outlines future research directions for AI voice agents in service and beyond.Design/methodology/approachDrawing on insights from marketing and management literature complemented with communications, linguistics, and human–computer interaction research, the article describes voice elements relevant for service research. The conceptual mapping and the organizing framework guide the future research agenda.FindingsFirst, AI voice is conceptualized through verbal (i.e. diction, syntax, semantics) and paraverbal elements (i.e. pitch, volume, speech rate, pauses, pronunciation, articulation, timbre, and breathiness). Next, the proposed typology maps AI voice agents by voice richness and AI intelligence. Finally, the conceptual framework articulates benefits (e.g. interactivity, social presence, convenience, personalization, enjoyment) and risks (e.g. intrusiveness, privacy concerns, algorithmic bias) associated with AI voice agents, and introduces relevant contingency factors covering user, voice technology, service setting, and ecosystem characteristics.Originality/valueThe paper offers a clear definition of AI voice, as well as a conceptual mapping that enables service researchers to systematically classify and compare AI voice agents. Moreover, the study highlights key pathways that may inspire future work
Data-Driven Robust Intertwined Wheat Supply Chain:Redesigning a viable network for long-term geopolitical conflicts
No full textThe recent human-made conflict in 2022 severely damaged Ukraine's infrastructure, causing significant instability in the food supply chain. This crisis was further exacerbated by trade bans imposed on another major global wheat exporter. Since wheat production and export are intrinsically linked, particularly in times of crisis, it is essential to adopt the concept of an intertwined supply chain. Accordingly, this study proposes an intertwined supply chain framework for the production and export of wheat during long-term disruptions. To enhance the viability of this intertwined system, the study introduces three key strategies. First, it addresses long-term disruptions and operational risks by employing redundancies and data-driven robust optimization techniques, where uncertainty sets are generated using a support vector clustering model. Second, the proposed supply chain accounts for freshwater resource limitations by integrating water resilience measures. Third, as the framework operates within a global context, it incorporates a comprehensive model that considers exchange rates, taxation, foreign demand points, and international trade responsibilities. To optimize these strategies, two multi-objective optimization models are developed and solved using an epsilon-constraint method. A cardinality-based measure is introduced to efficiently represent the Pareto front, offering decision-makers valuable insights into non-dominated solutions. The results are divided into analyses of wheat production, export, and their combined network. Individual analyses assess network setup, viability, and uncertainty control, while the integrated analysis examines sensitivity and interdependence. Overall, improving water use, managing risks, and designing a resilient, interconnected system can greatly strengthen the wheat supply chain during long-term crises
Efficient AI-aided segmentation of concrete CT scans using a predictive annotation interface
No full textConcrete is characterized by a heterogeneous mesostructure consisting of aggregates, mortar, voids, and the interfacial transition zone (ITZ). X-ray Computer Tomography (CT) image processing can capture the 3D mesostructure of concrete, enabling its reconstruction for generating numerical models. A crucial step in this process is the segmentation of CT images. However, common segmentation approaches, such as thresholding, are not feasible for concrete due to the overlapping grayscale values of aggregates and mortar, which make it impossible to define a threshold for their separation. To address this challenge, segmentation software tools based on trained neural networks have proven to be accurate and robust, especially in biomedical CT imaging. In this contribution, we describe a segmentation workflow for concrete CT images using the neural network-based software SuRVoS2. This workflow effectively segments the concrete mesostructure of unaltered samples into three phases: cement, aggregates, and voids. The segmentation procedure begins with a shallow learning step to achieve an initial segmentation of the volume with minimal manual labor. In the second step, a U-Net Convolutional Neural Network (CNN) is trained using subvolumes. The final step involves predicting the entire volume using the trained CNN model and results in accurate segmentation with F1-scores higher than 97%.}<br/
Soft robotics for personalized and sustainable wearables
No full textSoft robotic systems prevail in wearable and haptic applications, offering adaptability, safety and comfort. In this Review, we explore the integration of soft robotics in wearable designs for applications in assistance, rehabilitation and haptic sensory stimulation. We outline various types of soft actuators, examining their properties with regard to adjustable stiffness, mechanical responsiveness and sensing capabilities and their integration into wearable devices for health-care applications. We also highlight challenges and opportunities in developing sustainable, self-healing, self-powering and self-actuating soft robots, particularly with regard to achieving efficient energy usage, long-term durability and personalized control. Finally, we examine how machine learning might be explored to optimize the performance and adaptability of soft robotic devices to transform real-time data into actionable insights for personalized experiences.Soft robotic systems prevail in wearable and haptic applications, offering adaptability, safety and comfort. In this Review, we explore the integration of soft robotics in wearable designs for applications in assistance, rehabilitation and haptic sensory stimulation. We outline various types of soft actuators, examining their properties with regard to adjustable stiffness, mechanical responsiveness and sensing capabilities and their integration into wearable devices for health-care applications. We also highlight challenges and opportunities in developing sustainable, self-healing, self-powering and self-actuating soft robots, particularly with regard to achieving efficient energy usage, long-term durability and personalized control. Finally, we examine how machine learning might be explored to optimize the performance and adaptability of soft robotic devices to transform real-time data into actionable insights for personalized experiences
Surrogate Model-Based Reinforcement Learning for Bidding Strategies in Local Flexibility Markets
No full textReinforcement Learning (RL) has emerged as a promising tool for designing bidding strategies in electricity markets. Online RL learns policies by directly interacting with the environment, adapting in real time through trial-and-error feedback, while offline RL trains solely on a fixed, pre-collected dataset without any interaction with the environment. However, existing online RL methods assume detailed knowledge of the market-clearing model, including grid constraints, network parameters, and other participants' bids, which is unrealistic in a real-world setting. Additionally, the exploration process bears the risk of financial losses for market participants. Purely offline RL, on the other hand, suffers from its reliance on historical data that may be limited and non-diverse; as a result, it cannot gather new feedback to explore unobserved scenarios, potentially leading to sub-optimal results. To address these limitations, this paper proposes a surrogate model-based RL approach. Firstly, a machine learning surrogate model is constructed to approximate the local flexibility market (LFM) clearing process, using historical bids and corresponding market outcomes of the market participant. The surrogate model predicts the profit of the market participant associated with each bid in terms of price and quantity, thus avoiding the need for the true market model. A twin delayed deep deterministic policy gradient (TD3) RL bidding agent is then trained on this learned surrogate LFM model, enabling systematic exploration with reduced financial risk. Comparative simulations in the case study with three approaches—online RL, offline RL, and the proposed surrogate-based RL—reveal that online RL achieves the highest profit. Offline RL yields the lowest profit, constrained by limited historical data. The surrogatebased RL strategy outperforms the offline approach, offering a practical balance between profitability, and exploration risk
PCB-Based Hybrid Series/Corporate-Fed 4×4 D-Band Phased Array with Wide-Angle Scanning
No full textA wideband 4 × 4 planar phased array, based on four hybrid series/corporate-fed 4 × 1 subarrays with an asymmetrically positioned wide-stripline stub, is proposed for next-generation D-band applications. Each subarray consists of a stripline power splitter and 180 ◦ phase shifter serving as a corporate feed for two co-optimized back-to-back series-fed arrays, each with two stripline-fed aperture-coupled stacked patch antennas. This hybrid feed approach minimizes the number of required feed layers, reducing both structural complexity and fabrication costs, while ensuring stable gain across a wide bandwidth. As a proof of concept, prototypes of a 4 × 1 subarray and a 4 × 4 array, with respective dimensions of 0.96 mm × 4.8 mm and 3.8 mm × 4.8 mm, were fabricated using a standard any-layer high-density interconnect printed circuit board process for operation in the (120 to 150) GHz frequency band, demonstrating a 22% fractional bandwidth. The 4 × 1 and 4 × 4 arrays achieve peak gains of 10 dBi and 14.5 dBi, with total efficiencies of 58% and 55%, respectively, and support beam steering over \pm 45 ◦. Furthermore, a 5Gbaud 16-QAM wireless communication link over 26 cm distance was successfully established using the 4 × 4 array, achieving an error vector magnitude of less than 8%. This validates the use of the proposed antenna array at three carrier frequencies, enabling a 60 Gbit/s aggregated data rate for future multicarrier D-band applications.</p
The Life of Software Features: An Exploratory Case Study of 189 Feature Requests in Marlin
No full textFeatures are a widely established notion to organize the functionalities of a software system. For instance, features are used to define variability and commonalities in product lines; feature-driven development is an agile development methodology; and social-coding platforms have explicit support for feature requests. Despite the importance of features, we are not aware of extensive research on their life cycles: how and for what reasons do developers evolve features? As a result, we lack an understanding of how features come to be, how they are evolved, or why they may be removed. To narrow this research gap, we have performed an exploratory case study on the evolution of 189 feature requests of the Marlin 3D-printer firmware. We identified the code introducing a feature and traced all commits touching that code or the feature, resulting in a collection of 1,940 unique commits spanning five years of evolution. We have manually inspected all of these commits to classify their intentions with respect to the features they change, and created process graphs of the features’ life cycles based on these intentions to understand the evolution of features. Our results contribute a first overview and detailed examples of evolving features beyond code metrics, showcasing that features are primarily refactored, exhibit interdependent evolution, and are rarely removed. Serving as a starting point, these contributions can support practitioners in managing features and guide researchers in understanding feature evolution as well as in scoping future studies on this matter.</p
Exploring consecutive cycles of iron powder combustion for sustainable thermal energy
No full textTo advance the realization of iron powder as a clean energy storage and carrier, ten full combustion-reduction cycles of iron powder have been successfully demonstrated. Combustion was conducted using a lab-scale semi-practical metal powder burner, the "Metal Cyclonic Combustor (MC2)", while the reduction of iron oxide back to iron powder was achieved using hydrogen in a lab-scale semi-practical fluidized-bed reactor. The reduction process demonstrates high conversion, resulting in 87±0.8 wt% of iron content in the recycled iron powder. The recycled iron powder achieves a combustion efficiency of 85±0.9% compared to the theoretical maximum heat release of iron to hematite, which is comparable to the 86% efficiency of virgin iron powder. It also maintains stable combustion characteristics over ten cycles. Despite slightly lower gas temperatures due to residual iron oxides in the recycled powder, the combustion process remains stable, with consistent iron flames and uniform gas temperatures. The stability in combustion performance is attributed to the preserved particle size of recycled iron powder and its porosity, which facilitates ignition. Furthermore, emissions, including nanoparticle and NO formation, remain constant throughout the cycles, with NO emissions below 2.5 mg/MJ, which is significantly lower than those of other combustion fuels. These promising results highlight the feasibility of iron powder as a clean and safe energy storage and carrier medium, offering a lower-risk alternative to conventional fuels.</p
Integrating Flow Field Geometries within Porous Electrode Architectures for Enhanced Flow Battery Performance
No full textThe large-scale adoption of renewable energy demands efficient and cost-effective storage solutions, with redox flow batteries (RFBs) emerging as promising candidates for grid-scale applications. However, their deployment remains constrained by high capital costs, largely driven by the need for advanced porous electrodes that balance high surface area, efficient mass transport, and low-pressure drop. Compared to conventional, carbon-fiber-based porous electrodes, non-solvent induced phase separation (NIPS) offers a versatile manufacturing approach to tailor electrode microstructures and enhance electrochemical performance, yet optimizing mass transport remains a key challenge. Here, a micro-patterning strategy is introduced that directly integrates flow field architectures into the electrode structure during NIPS fabrication as a potentially scalable manufacturing approach. Inspired by flow field designs used in fuel cells and flow batteries, we imprint groove and pillar micro-patterns to enhance in-plane and through-plane mass transport. Using symmetric iron flow cells and all-vanadium full cells, pillar-patterned electrodes, combined with an interdigitated flow field, are shown to significantly reduce mass transfer resistance and improve electrochemical performance while maintaining a low-pressure drop. This work presents a simple, scalable, and cost-effective electrode design strategy to boost RFB power density and advance the economic viability of redox flow battery technology.</p