275003 research outputs found
Sort by
Transfer Learning in Multi-Agent Reinforcement Learning with Double Q-Networks for Distributed Resource Sharing in V2X Communication
No full text146151This paper addresses the problem of decentralized spectrum sharing in vehicle-to-everything (V2X) communication networks. The aim is to provide resource-efficient coexistence of vehicle-to-infrastructure (V2I) and vehicle-to-vehicle (V2V) links. A recent work on the topic proposes a multi-agent reinforcement learning (MARL) approach based on deep Q-learning, which leverages a fingerprint-based deep Q-network (DQN) architecture. This work considers an extension of this framework by combining Double Q-learning (via Double DQN) and transfer learning. The motivation behind is that Double Q-learning can alleviate the problem of overestimation of the action values present in conventional Q-learning, while transfer learning can leverage knowledge acquired by an expert model to accelerate learning in the MARL setting. The proposed algorithm is evaluated in a realistic V2X setting, with synthetic data generated based on a geometry-based propagation model that incorporates location-specific geographical descriptors of the simulated environment (outlines of buildings, foliage, and vehicles). The advantages of the proposed approach are demonstrated via numerical simulations
Large deflection of electrostatically actuated microbeams – How do the bending modes contribute beyond pull-in?
No full text407410Coulomb-actuated microbeams are widely used in micro-electro-mechanical systems. Current technological trends, such as 5G Internet of Things (5G-IoT), augmented reality, and Green ICT (information and communications technology), drive system level modelling, which requires accurate and efficient lumped parameter models (LPM) of single electromechanical components (e.g. MEMS sensors). However, since electro-mechanical systems are highly nonlinear in nature, the development of accurate models is challenging. For actuated beams, a state of the art analytical technique is to expand the beam deflection in terms of suitably chosen eigenfunctions. The targeted LPMs are then obtained by projecting the governing differential equations onto the low dimensional subspace spanned by those eigenmodes (Galerkin projection), deemed relevant. LPMs obtained this way, are hitherto often computationally expensive because, they typically require several degrees of freedom in order to achieve the desired accuracy at larger deflections. The work presented here solves this issue for prismatic Coulomb activated Euler-Bernoulli beams, clamped at both ends. The actual contributions of the leading Euler-Bernoulli eigenmodes to the deflection profiles are investigated, numerically and experimentally. We find that the lowest Euler-Bernoulli mode by far dominates the physics of these devices (zero-mode hypothesis). The finding holds beyond the pull-in, up to a small vicinity of the contact singularity. As a result we are able to develop a new LPM, substantially improving the state of the art, with respect to accuracy and computational simplicity. This paper is a synopsis of Refs. [1] and [2]
Capacitive Micromachined Ultrasonic Transducers for Applications in Nondestructive Testing
No full text676679Capacitive Micromachined Ultrasonic Transducers (CMUTs) are characterized by a high sensitivity in receiving and a wide bandwidth based by their physical properties. The development of these sensors was primarily driven by medical diagnostics in the last decades. CMUTs are best suited for the classical sonography because of their natural matching to water. The good adaptability to air additionally enables the application at diagnostic procedures, which do not allow the use of coupling liquids, for example on the eyeball or in the ear canal. The improvement in the optimization of the CMUT structures at the Fraunhofer IPMS enables the addressing of further fields of application such as the Nondestructive Testing (NDT). CMUTs are suitable especially for the implementation of small apertures and arrays with a large number of elements with up to 30 MHz working frequency. The advantages in comparison with piezoelectric transducers come into play particularly at imaging methods. We give an overview of the advantages of CMUT probes and show the opportunities in their application for NDT with examples. Further, we present the available ultrasonic transducers and demonstrator devices made by Fraunhofer IPMS
Illusion of Precision: How Averaging Undermines Pulse Measurement Accuracy
No full text222233Accurate heart rate measurement is a fundamental requirement in many fields, including medicine, veterinary science, sports performance, and health monitoring. Devices designed to monitor pulse rate are widely used in both human and animal applications, and their reliability is often taken for granted.
However, a curious and counterintuitive effect can occur: some devices consistently report a minor changing value - such as 72 beats per minute for heart beat detection - regardless of the actual physiological state of the subject. At first glance, this appears to be a minor flaw, especially since the reported value may align with the average resting heart rate in humans. From a statistical standpoint, the device could even appear reasonably accurate when evaluated over a large group of individuals. Yet this illusion of accuracy masks a critical flaw: individual deviations are ignored, and potentially important physiological variations go undetected. This paper explores how averaging can give rise to the appearance of precision, while in reality, the measurement fails to reflect real-time or individualized heart rate data. We argue that devices with such behavior introduce systematic bias, which poses significant risks in both clinical and practical applications
Chronic Spontaneous Urticaria and Comorbidities
No full text77107Chronic spontaneous urticaria (CSU) is a common and debilitating disease, which frequently coexists with other diseases, i.e., comorbidities. Some comorbidities, e.g., autoimmune diseases, are more common in CSU patients, and CSU patients may benefit from screening for these conditions. Other CSU comorbidities such as metabolic syndrome are under-recognized in CSU. As of yet, urticaria guidelines do not provide specific diagnostic recommendations on screening for these comorbidities. In addition, CSU comorbidities, for instance, mental disorders and chronic inducible urticarias (CIndUs), add to the burden of disease and quality of life impairment. Their identification and treatment, in clinical practice, can help to optimize the management of CSU patients. Furthermore, some diseases, namely CIndUs and Hashimoto’s thyroiditis, were suggested to be markers of longer CSU duration and progression from acute spontaneous urticaria to CSU. Finally, certain diseases are linked, pathogenetically, to CSU, although the mechanisms are yet to be defined. Investigation of these diseases can help to better understand CSU, and their treatment can reduce CSU disease activity. In particular, some case reports supported this notion showing CSU remission or improvement after the treatment of malignancy, infection, and hyper- and hypothyroidism. In this chapter, we describe important groups of comorbidities of CSU, their prevalence, and their relevance for clinical practice
Enhanced Audio-Visual Speech Synthesis Via Multi-Discriminative Learning
No full textAudio-visual speech synthesis (AVSS) aims to produce an audio-visual stream that conveys a target speaker's speech. In this study, the AVSS system takes the input speech of a source speaker and generates the audio-visual stream of the target speaker while preserving the linguistic content of the source speech. The process involves two main components: voice conversion (VC), which adapts the vocal features from the source to the target speaker, and audio-visual synthesis (AVS), which generates the synchronized audio-visual stream from the transformed speech. This paper presents a novel generative framework based on multi-discriminative learning to enhance the realism and quality of AVSS outputs. The proposed approach integrates multiple discriminators, including capsule networks, co-occurrence neural networks, and vision transformers (ViTs), within the VC model to leverage their unique strengths in capturing diverse speech features. Additionally, the AVS model incorporates a co-occurrence neural network to improve video quality and achieve better temporal alignment between audio and visual data. Experimental evaluations on standard benchmarks demonstrate that the proposed method achieves significant improvements in both audio and video quality, offering a substantial advancement in AVSS technology.Online Firs
Advancing the prediction of factors associated with bipolar disorder risk: utilizing early recognition tools and polygenic risk scores
No full textBipolar disorder (BD) is a highly heritable mental illness that affects ∼ 1-2% of the world’s population and has complex genetic and environmental underpinnings. Early detection is critical to improving treatment outcomes, but current strategies have limited predictive power. Early detection tools such as the Early Phase Inventory for Bipolar Disorder (EPIbipolar) and the Bipolar At-Risk (BARS) criteria assess phenotypic risk factors, including family history (FH) and subthreshold mood problems. Polygenic risk scores (PRS) are a quantitative metric of genetic susceptibility. This study examined the associations between BD-PRS and screening tools in order to assess their combined potential to identify individuals at risk of BD with improved predictive accuracy. The analysis included 1068 participants, including 199 at-risk young adults aged 15 to 35 years and 869 healthy controls aged 18 to 50 years. All of them had no prior psychiatric disorders. Inclusion criteria for the at-risk group comprised a positive FH (1st or 2nd degree) for BD, major depressive disorder (MDD), attention-deficit/hyperactivity disorder (ADHD), or the presence of specific BD risk factors (e.g., subthreshold hypomanic symptoms, mood swings, or sleep disturbances). Participants who had a confirmed BD, schizophrenia, schizoaffective disorder diagnosis, or other psychiatric conditions that could explain the symptomatology, were excluded. Diagnostic assessments that were utilized validated early detection instruments, including EPIbipolar, Bipolar Prodrome Interview and Symptom Scale-Prospective (BPSS-FP), and BARS criteria. Binary logistic regression models were employed to assess associations between BD-PRS and phenotypic risk markers, with adjustments for population stratification. Results revealed significant associations between BD-PRS and BARS criteria risk groups and EPIbipolar "at risk" criteria compared to controls. Significant associations were also identified for subscales including FH for BD, MDD, or schizophrenia, sleep and circadian rhythm disturbances, depressive characteristics, functional impairment, and episodic course. However, no significant associations were observed between BD-PRS and BPSS-FP, which highlights variability in the sensitivity of different early detection instruments. Our findings emphasize the potential of combining genetic susceptibility measures with phenotypic risk markers to enhance early detection strategies for BD. Further research is needed to optimize predictive models and evaluate the clinical utility of PRS in early intervention frameworks.14
ATLAS livestock monitoring architecture and services
No full text10221029The continuous monitoring of livestock behaviour is an important task in precision livestock farming. An early recognition of critical events in a herd can lead to an improved animal welfare and thus for example in a reduced need of antibiotics. Furthermore, it has been found that even feeding intake and efficiency can be affected by animal behaviour. Due to the size of modern livestock farming operations, the usage of automated systems becomes necessary. We present a scalable automated monitoring system composed of low-cost IP video surveillance cameras and affordable edge-computing hardware on-site, complemented by standardized web services that allow for long term video storage and analysis. For the automated analysis of eating- and resting time as well as activity levels in videos, we perform state-of-the-art object detection and tracking using a YOLOv3 DeepSORT Convolutional Neural Network. Data management and the integration of the analysis system into the farmers’ work-flows are achieved using a distributed service interoperability network that allows the exchange of data between different services related to livestock management. The interoperability network is completely federated and based on established web technologies. Interoperability is achieved through standardized APIs and data formats. This enables farmers for example to seamlessly integrate behaviour analysis results into feeding management systems, or to share the results with external consultancy services or other stakeholders in the production chain. Furthermore, the standardized data exchange removes the extra workload that is caused by entering the same data into different systems
The mechanism of macroautophagy: The movie
No full text414417This animated movie presents the mechanism of macroautophagy, hereafter autophagy, by showing the molecular features of the formation of autophagosomes, the hallmark organelle of this intracellular catabolic pathway. It is based on our current knowledge and it also illustrates how autophagosomes can recognize and eliminate selected cargoes.1
Acoustics for monument preservation - Measures, planning and validation at three real-world laboratories
No full text334341Die Optimierung der Raumakustik in denkmalgeschützten Gebäuden erfordert den Erhalt historischer Substanz bei gleichzeitig kurzen Nachhallzeiten für Sprache und Musik. Zu diesem Zweck wurden im Fraunhofer IBP im Rahmen eines Projektes drei Reallabore (Forum Heimat, Franz‐Marc‐Raum, Großer Sitzungssaal Deichmannsaue) ausgewählt, in denen reversible Akustikmaßnahmen erprobt und validiert wurden. Nach einer systematischen Literaturrecherche und zehn leitfadengestützten Experteninterviews wurden die Räume in SketchUp 3D‐modelliert, in ODEON simuliert und mittels Vor‐Ort‐Nachhallzeitmessungen kalibriert. Iterativ wurden Nischenabsorber (22 m) analysiert. In den Reallaboren ließen sich Nachhallzeiten auf A2‐Niveau gemäß DIN 18041 reduzieren. Die In‐situ‐Messungen zeigten deutliche Abweichungen zu Hallraumdaten, weshalb eine Feinabstimmung erforderlich ist. Ein kompaktes Planungskompendium fasst praxisnahe Empfehlungen für Architekt:innen und Planer:innen zusammen.Heritage-listed buildings pose a unique challenge for room acoustics: preserving historic fabric while achieving low reverberation times for speech and music. To address this, three real-world laboratories (Forum Heimat, Franz-Marc Room, large conference hall at Deichmannsaue) were selected as part of a project at Fraunhofer IBP to test reversible treatments. Following a structured literature review and ten expert interviews, each space was 3D-modelled in SketchUp, simulated in ODEON, and calibrated with on-site reverberation time measurements. Iterative evaluations of niche absorbers (22m2 Reapor), transparent micro-perforated Plexiglas column claddings (5.5m2) and mobile partitions (120m2) reduced reverberation to meet DIN 18041 A2 criteria. On-site tests revealed significant deviations from reverberationchamber data, highlighting the need for fine-tuning. A concise planning compendium synthesizes actionable guidance for architects and acousticians.47