31325 research outputs found
Sort by
Model‐Driven Engineering for Digital Twins: Opportunities and Challenges
No full textDigital twins are increasingly used across a wide range of industries. Modeling is a key to digital twin development—both when considering the models which a digital twin maintains of its real‐world complement (“models in digital twin”) and when considering models of the digital twin as a complex (software) system itself. Thus, systematic development and maintenance of these models is a key factor in effective and efficient digital twin development, maintenance, and use. We argue that model‐driven engineering (MDE), a field with almost three decades of research, will be essential for improving the efficiency and reliability of future digital twin development. To do so, we present an overview of the digital twin life cycle, identifying the different types of models that should be used and re‐used at different life cycle stages (including systems engineering models of the actual system, domain‐specific simulation models, models of data processing pipelines, etc.). We highlight some approaches in MDE that can help create and manage these models and present a roadmap for research towards MDE of digital twins.Deutsche Forschungsgemeinschaft 10.13039/501100001659Agence Nationale de la Recherche 10.13039/501100001665European Commission 10.13039/501100000780Peer Reviewe
Unknown Germany - An integrative biodiversity discovery program
No full textBiodiversity knowledge, from genes to ecosystems, is crucial for addressing the biodiversity crisis. However, even in well-explored countries like Germany, much biodiversity remains unknown. Therefore, several research institutions are joining forces to conduct a comprehensive biodiversity inventory, combining broad taxonomic expertise with advanced technologies. By consolidating data across many organismic groups, the Unknown Germany initiative will significantly enhance conservation strategies and may serve as a model for similar efforts worldwide.Peer Reviewe
Knowledge-augmented pre-trained language models for biomedical relation extraction
No full textAutomatic relationship extraction (RE) from biomedical literature is critical for managing the vast amount of scientific knowledge produced each year. In recent years, utilizing pre-trained language models (PLMs) has become the prevalent approach in RE. Several studies report improved performance when incorporating additional context information while fine-tuning PLMs for RE. However, variations in the PLMs applied, the databases used for augmentation, hyper-parameter optimization, and evaluation methods complicate direct comparisons between studies and raise questions about the generalizability of these findings. Our study addresses this research gap by evaluating PLMs enhanced with contextual information on five datasets spanning four relation scenarios within a consistent evaluation framework. We evaluate three baseline PLMs and first conduct extensive hyperparameter optimization. After selecting the top-performing model, we enhance it with additional data, including textual entity descriptions, relational information from knowledge graphs, and molecular structure encodings. Our findings illustrate the importance of (1) the choice of the underlying language model and (2) a comprehensive hyperparameter optimization for achieving strong extraction performance. Although inclusion of context information yield only minor overall improvements, an ablation study reveals substantial benefits for smaller PLMs when such external data was included during fine-tuning.Open Access funding enabled and organized by Projekt DEAL.Humboldt-Universität zu Berlin (1034)Peer Reviewe
Journal hijacking: challenges for the scientific community and recommendations for journals
No full textPeer Reviewe
Efficient 6D Pose Estimation for Augmented Reality through Hybrid Methods and Domain-Invariant Matching
No full textIn Augmented Reality-Anwendungen sind präzise Kamera- und Objektposen notwendig, um virtuelle Inhalte stabil in die reale Umgebung einzubetten. Trotz Fortschritten durch tiefe neuronale Netze bleibt der Abgleich synthetischer Referenzdaten mit realen Kamerabildern herausfordernd. Der Reality Gap entsteht durch Unterschiede in Darstellung und Annahmen zwischen synthetischen und realen Daten. Diese Dissertation entwickelt Verfahren zur sechsdimensionalen Posenschätzung von Kameras und Objekten, die domäneninvariante Merkmale wie Silhouetten, Geometriekanten, semantische Layouts und projizierte Texturen nutzen, um den Abgleich zwischen synthetischen und realen Daten zu verbessern. Durch die Kombination lern- und modellbasierter Ansätze wird der Reality Gap reduziert und die Generalisierung auf neue Geometrien erleichtert. Ein neues Verfahren zur projektorbasierten 6D-Posenverfeinerung für Spatial Augmented Reality berücksichtigt Verzerrungen projizierter Texturen bei der Analyse durch Synthese. Die Projektion, die bestehende bildbasierte Systeme stört, dient hier der Registrierung. Ein neuer Ansatz zur 6D-Posenschätzung erhöht die Kapazität eines Convolutional Neural Networks für mehrere Objektklassen, indem Klassen- und Forminformationen aus semantischen Masken genutzt werden, um 2D-3D-Korrespondenzen einstufig zu schätzen. Ein hybrider Ansatz kombiniert KI-basierte Posenschätzung mit lokaler Verfeinerung und nutzt die 3D-Geometrie zur Fehlererkennung. Dies vereinfacht das Training und steigert die Robustheit in Echtzeitanwendungen, besonders bei geometrisch ähnlichen Objekten. Schließlich wird ein Verfahren zur 6D-Kameraposenschätzung in Innenräumen entwickelt, das Panoramarenderings eines semantischen 3D-Modells als Referenz nutzt, um globale Posen in Szenen zu bestimmen, die im Training ungesehen waren. Zahlreiche Experimente zeigen die Leistungsfähigkeit und Vorteile der Methoden. Praktische Beispiele verdeutlichen ihre Anwendbarkeit im Augmented Reality-Kontext.In augmented reality applications, precise camera and object poses are necessary to stably embed virtual content into the real environment. Despite advances through deep neural networks, aligning synthetic reference data with real camera images remains challenging. The reality gap arises from differences in representation and assumptions between synthetic and real data. This thesis develops new methods for six-dimensional pose estimation of cameras and objects, using domain-invariant features such as silhouettes, geometric edges, semantic layouts, and projected textures to improve alignment between synthetic and real data and enhance pose estimation accuracy. By combining learning-based and model-based approaches, the reality gap is reduced, and generalization to new geometries is facilitated. A novel method for projector-based 6D pose refinement in spatial augmented reality accounts for distortions of projected textures through analysis-by-synthesis. The projection, which disrupts conventional image-based systems, is used here for registration. A new 6D pose estimation approach increases the capacity of a convolutional neural network for multiple object classes by utilizing class and shape information from semantic masks to estimate 2D–3D correspondences in a single stage. A hybrid approach combines AI-based pose estimation with local refinement and uses 3D geometry for error detection. This simplifies training and increases robustness in real-time applications, especially for geometrically similar objects. Finally, a method for 6D camera pose estimation in indoor environments is presented, using panorama renderings from a semantic 3D model as a reference to determine the global pose in scenes not seen during training. Numerous experiments demonstrate the performance and advantages of the methods. Practical examples illustrate their effectiveness in augmented reality applications
Impacts of fertilizers coupled with improved seeds on rice profitability in Tanzania: evidence from the national sample census of agriculture 2019/20
No full textIntroduction: Climate change poses a growing threat to food security in Tanzania, particularly among smallholder rice farmers who rely on rain-fed agriculture and have limited adaptive capacity. Identifying effective input strategies to enhance both profitability and resilience is crucial to securing rural livelihoods and achieving sustainable development goals. This study evaluates the potential of improved seed varieties and fertilizers (both organic and inorganic) as tools for boosting income and reducing vulnerability among rice farming households. Methods: We utilized nationally representative data from the 2019/20 National Sample Census of Agriculture (NSCA), which encompassed 6,025 rice farms across diverse agroecological zones in Tanzania. A Risk Simulation Model (RSM) was employed to estimate farm income distributions under varying yield and market price conditions. Farmers were categorized by seed type (local vs. improved) and fertilizer use (none, organic, or inorganic). Income thresholds of TZS 2.0 million/ha (low) and TZS 4.0 million/ha (high) were used to benchmark profitability and vulnerability levels across regions. Results: The combination of improved seeds and inorganic fertilizers significantly enhanced farm profitability. In the Southern Highlands Zone (SHZ), 20% of farms exceeded the upper-income threshold of TZS 4.0 million/ha. In contrast, farmers using local seeds without fertilizer faced a high financial risk, with 78% earning less than TZS 2.0 million/ha nationally. Organic fertilizers offered modest gains; in Zanzibar, they reduced the probability of falling into the low-income category by 14%. However, regional disparities were substantial. Fertile soils in the SHZ amplified input returns, while constraints in the Eastern Zones (EZ) and Zanzibar, including saline soils, land fragmentation, and institutional limitations, restricted profitability, even among users of inorganic fertilizer, with 58% remaining below the income threshold. Discussion: These findings underscore the importance of spatially targeted strategies. High-potential areas, such as SHZ, could benefit from scaling up access to improved inputs through subsidized programs, like digital vouchers. In low-performing regions like EZ and Zanzibar, integrated interventions are needed, including site-specific soil management, saline-tolerant seeds, infrastructure improvements, and access to credit. A dual policy approach, combining short-term subsidies with long-term resilience-building, is essential to advance food security and sustainable profitability, in line with SDG 2 (Zero Hunger).Peer Reviewe
MOF-based catalysts for sustainable biodiesel production: classification, performance, and advances from 2020 to 2025
No full textGlobal energy demand and environmental concerns have intensified the search for renewable fuels, with biodiesel emerging as a sustainable substitute for petroleum diesel. Efficient catalysis remains the bottleneck for large-scale biodiesel production. While heterogeneous catalysts offer advantages of reusability and separation, their performance is limited by stability, active sites availability, and reduced activity under harsh conditions. Metal–organic frameworks (MOFs), with their high surface area, tunable porosity, and structural versatility, have recently attracted increasing attention as next-generation catalysts. This work reviewed advances in the design and application of each common MOF type for biodiesel synthesis through esterification and transesterification process. MOF composites, MOF derivatives, and MOF composite materials exhibit superior catalytic performance and recyclability compared to pristine MOFs, making them highly recommended for future research and applications. Beyond summarizing yields and reaction conditions, we highlight mechanistic insights, stability issues, and catalysis performance. Special attention is given to functionalized and composite MOFs, bifunctional systems, and enzyme-MOF hybrids, which demonstrate superior performance compared to pristine MOFs. While UiO- and ZIF-based MOFs dominate current research, emerging systems such as Ca- and Cu-MOFs remain underexplored yet promising. We analyze the key features required in MOF materials for efficient biodiesel production and provide a comprehensive review and categorization of recent advancements. By contrasting MOFs with conventional heterogeneous catalysts and positioning this review against existing literature, we provide a comprehensive and critical perspective on the opportunities and challenges of MOFs in biodiesel catalysis.Deanship of Scientific Research, Imam Mohammed Ibn Saud Islamic University 10.13039/501100023673Peer Reviewe
Creative experiences and brain clocks
No full textCreative experiences may enhance brain health, yet metrics and mechanisms remain elusive. We characterized brain health using brain clocks, which capture deviations from chronological age (i.e., accelerated or delayed brain aging). We combined M/EEG functional connectivity ( N = 1,240) with machine learning support vector machines, whole-brain modeling, and Neurosynth metanalyses. From this framework, we reanalyzed previously published datasets of expert and matched non-expert participants in dance, music, visual arts, and video games, along with a pre/post-learning study ( N = 232). We found delayed brain age across all domains and scalable effects (expertise>learning). The higher the level of expertise and performance, the greater the delay in brain age. Age-vulnerable brain hubs showed increased connectivity linked to creativity, particularly in areas related to expertise and creative experiences. Neurosynth analysis and computational modeling revealed plasticity-driven increases in brain efficiency and biophysical coupling, in creativity-specific delayed brain aging. Findings indicate a domain‑independent link between creativity and brain health.Peer Reviewe
Local Fermi Level Engineering in 2D-MoS2 Realized via Microcontact Printing of Self-Assembled Monolayers for Next-Generation Electronics
No full textThe article processing charge was funded by the Open Access Publication Fund of Humboldt-Universität zu Berlin.Silicon-based technology is approaching scalability limits due to severe short-channel effects arising from its intrinsic bulk properties. In contrast, two-dimensional (2D) transition metal dichalcogenides (TMDCs) exhibit remarkable resilience to these effects because of their atomic-scale thickness, positioning them as promising candidates for next-generation optical and electronic devices. However, realizing 2D material-based technology still requires the development of local p- and n-type doping methods essential for complementary circuits. Self-assembled monolayers (SAMs) have shown the ability to locally engineer electronic energy levels in 2D TMDCs to address this challenge. In this study, we demonstrate local engineering of electronic energy levels on micrometer scale in semiconducting single-layer (1L) MoS2 by patterning the supporting substrate with functional SAMs via microcontact printing (µCP). Three SAMs were selected: two with large opposing dipole moments and one non-dipolar reference. Their impact on surface properties particularly the work function and on optoelectronic properties of 1L-MoS2 was investigated via Kelvin probe microscopy and photoluminescence (PL) mapping. Significant shifts in work function and PL were observed. FETs fabricated on locally patterned substrates enabled direct comparison, confirming that threshold voltage shifts up to 80 V and ON-current increases by two orders of magnitude arise solely from SAM polarity. This work demonstrates that µCP and the electrostatic doping capabilities of dipolar SAMs offer a straight forward and scalable approach to locally engineering 1L-MoS2 energy levels.Peer Reviewe