154092 research outputs found
Sort by
Multi-Axis Large-Range Silicon-Micromachined Force Sensing System
No full textThis paper reports the first microfabricated modular multi-degree-of-freedom (DOF) force sensing system using a p+ silicon and glass wafer stack. The proposed 2-DOF sensor combines a large measurement range and small 2×2 mm2 active region with the use of common and off-the-shelf substrates. The four mask fabrication method entails dry-etching of electrodes and 300 µm high micro-pillars into bulk silicon, which are anodically bonded to a glass wafer containing micromachined cavities with gold electrodes. Four 2-DOF sensor elements can be assembled in a single package to allow for larger-range 6-DOF force/torque sensing functionality. The sensor was characterized for 0–15 N in normal direction and 0-2.5 N in shear direction, showing relatively high sensitivities for the sensor size due to the dense electrode array
Simultaneous Perturbation of Knee and Ankle Joints during Stance Phase of Walking:Towards Biological Joint Impedance Estimation
No full textCharacterization of lower limb joint impedance during locomotion is crucial for advancing our understanding of walking biomechanics, while also guiding the development of assistive technologies and rehabilitation strategies. Identification of impedance parameters is dependent on perturbation strategies which elicit measurable responses at the joint of interest. However current perturbation methodologies applied during locomotion either work in the swing phase or employ exoskeleton based setups which can add complexity and weight to the leg. This study introduces and evaluates a perturbation methodology using a custom-designed pusher system to elicit multi-joint-level torque and angular responses during the stance phase of gait for different conditions. Controlled perturbations applied behind the tibia revealed measurable dynamic response at the knee and ankle joint, while preserving the gait cycle's overall pattern. Torque-angle and torque-velocity graphs reveal insights into energy and power changes between perturbed and unperturbed gait cycles. The method can be adapted to various walking speeds, perturbation intensities, and durations, providing the first step for future estimation of joint impedance during locomotion.</p
Understanding factors influencing ride-splitting adoption in Beijing: A comparative analysis with solo ride-hailing
No full textRide-splitting, a special kind of ride-hailing service, presents significant potential for energy savings and emission reduction. Studying factors that promote ride-splitting can help build sustainable transportation systems. Although many studies have analyzed the impact of the built environment and sociodemographic variables on ride-splitting, there is a lack of consideration of variables specific to ride-hailing systems. This study aims to analyze the complex impact of explanatory variables (including ride-hailing system-specific variables) on ride-splitting, based on an interpretable machine-learning framework. Firstly, the price ratio between shared and solo trips, the distance passengers wait for the driver to pick them up (called passenger waiting distance), and the driver’s detour index are extracted from Beijing’s data. Then, a machine learning-based framework combining XGBoost and SHAP is constructed. The explained variables are the daily trip numbers of ride-splitting and solo ride-hailing between origin–destination (OD) pairs. The results show that price ratio, passenger waiting distance, and detour index have a greater impact on ride-splitting than solo ride-hailing. Based on SHAP values, a nonlinear threshold-based relationship between individual variables and ride-splitting demand is investigated. Exogenous variables related to the high adoption of ride-splitting include OD pairs having trip durations shorter than 20 min, a zonal per capita GDP below a certain threshold, and being located away from the city center. The interaction effects of multiple variables on ride-splitting, such as distance from the origin/destination to the city center and travel time, are investigated based on the SHAP interaction value. These findings help to adapt specific variables to facilitate the shift from solo trips to shared trips, which is conducive to more sustainable transportation patterns
Ecological Validity of Grief Rumination Measures Among Bereaved People
No full textGrief rumination, characterised by repetitive thinking about the loss and its causes and consequences, is linked to various psychopathological symptoms, including prolonged grief disorder (PGD). Traditional assessments of grief rumination rely on trait self-report questionnaires assessing multiple types of rumination (e.g., reactions, injustice, counterfactuals and meaning), which may be susceptible to memory biases and often fail to capture the dynamic and context-dependent nature of ruminative thoughts. This brief report evaluates the ecological validity of trait versus state grief rumination types using ecological momentary assessment (EMA), testing their convergent and discriminant validity. Bereaved adults (N = 65, 42 women, Mage = 21.88 ± 2.92) completed online measures for 11 types of trait and state grief rumination. The state measures were completed four times a day for 14 consecutive days. We examined the convergent and discriminant validity of these measures through zero-order and multivariate associations. The associations between trait and state grief rumination measures varied between 0.388 and 0.765, but we did not find sufficient evidence supporting the discriminant validity. Multilevel regression analyses further indicated that trait measures of grief rumination captured a fraction of the individual's state grief ruminations. Our findings suggest that trait grief rumination measures may not fully capture the nuances of grief rumination experienced in daily life after loss. We therefore recommend using state measures given they more accurately seem to assess the ebb and flow of grief rumination in real-world settings.</p
Challenges in Adoption and Scaling of AI:A Case Study at a High-Tech Firm and Research Roadmap
No full textThe rapid evolution of AI has created a wealth of possibilities for businesses, e.g. to enhance their operations, improve their innovativeness and advance customer interaction. Technology departments are under pressure to adopt AI in their enterprise architecture. The immense power and readily available services offered by big tech platforms seems make design, implementation and operation of AI applications, including machine learning and deep learning, seamless. The paradigm of Machine Learning Operations (MLOps) emerged to develop ML products and rapidly bring them into production at industrial scale. It has been found that DevOps teams can contribute to firm competitive advantage by building both business and technology-related capabilities which enable them to sense market opportunities, make fast and targeted decisions and transform their assets in case of changing circumstances. While increasingly popular, MLOps has shown to be difficult. Many ML initiatives fail to provide value, while many ML models never reach production. This study surveys challenges of AI adoption and discusses a framework based approach to facilitate the adoption and scaling of AI in a tech firm. The paper concludes that while a framework based approach does eliviate some adoption challenges, much research remains to be done. Such research challenges are presented and discussed.</p
Prostate Volume and PSA-Density Estimation by Transabdominal Ultrasound:Prospective Evidence of Comparative Accuracy to MRI and Transrectal Ultrasound in Prostate Cancer Early Diagnostics
No full textBackground: Prostate-specific antigen (PSA) density is an easily available predictor for clinically significant PCa. While transrectal ultrasound (TRUS) is utilized for PSA density (PSAD) estimation, transabdominal ultrasound (TAUS) is a more accessible, noninvasive alternative that can be used to decide if follow-up diagnostics are necessary. This study aims to compare prostate volume (PV) and PSAD across TAUS, TRUS and MRI, comparing the clinical utility of TAUS and TRUS for PSAD-based risk stratification. Methods: Hundred and eighty men undergoing PCa diagnostics were included by collecting serum PSA, TRUS, MRI, and TAUS PV examinations. PV was calculated blindly by all operators and image quality was assessed. Agreement in PV measurements of all imaging modalities was analyzed in Bland-Altman diagrams. PCa risk derived from PSADTAUS and PSADTRUS was compared against MRI outcomes in Sankey diagrams and the percentage of misclassified PCa risk was reported. Results: After excluding 33 inadequate TAUS acquisitions, 147 patients were included. The average volume difference between TAUS and MRI was 2.5 mL (standard deviation (SD): 16.4), between TAUS and TRUS 11.5 mL (SD: 20.4), and between TRUS and MRI −9.0 mL (SD: 21.1). TAUS and TRUS underestimate PCa risk in 3%–4%, while the percentage of men with overestimated risk decreased when TAUS was used (7% vs. 13%). Conclusions: PVs obtained with TAUS are equivalent to MRI. Still, image quality varies with experience and interobserver variability needs further exploration, ensuring generalizable outcomes. Nevertheless, TAUS represents a valid alternative for PV and PSAD estimation, enabling a patient-friendly alternative for PCa risk assessment.</p
Model Order Reduction for Seismic Applications
No full textIn this paper, we propose a model order reduction approach to speed up the computation of seismograms, i.e., the solution of the seismic wave equation evaluated at a receiver location, for different model parameters. This is highly relevant for seismic applications, such as full waveform inversion, seismic tomography, or monitoring tools of seismicity, that are computationally challenging,as the discretized (forward) model often has a huge number of unknowns and needs to be solved many times for different model parameters. Our approach achieves a reduction of the unknowns by a factor of approximately 1000 for various numerical experiments for a two-dimensional subsurface model of Groningen, the Netherlands (a region known for its seismic activity), even if the (known) wave speeds of the subsurface are relatively varied. Moreover, when using multiple cores to constructthe reduced model, we can approximate the (time domain) seismogram in a lower wall-clock time than an implicit Newmark-beta method. To realize this reduction, we exploit the fact that seismograms are low-pass filtered for the observed seismic events. We thus consider the Laplace-transformed problem in the frequency domain and implicitly restrict ourselves to the frequency range of interest by adjusting the parameters of a source function, which is a popular model in computational seismology for the temporal response of an earthquake. Therefore, we can avoid the high frequencies that would require many reduced basis functions to reach the desired accuracy and generally make the reduced order approximation of wave problems challenging. Instead, we can prove for our ansatz that for a fixed subsurface model, the reduced order approximation converges exponentially fast in thefrequency range of interest in the Laplace domain. We build the reduced model from solutions of the Laplace-transformed problem via a (proper orthogonal decomposition-)greedy algorithm targeting the construction of the reduced model to the time domain seismograms; the latter is achieved by using an a posteriori error estimator that does not require computing any time domain counterparts.Finally, we show that we obtain a stable reduced model, thus overcoming the challenge that standard model reduction approaches do not necessarily yield a stable reduced model for wave problems
Estimating soil organic carbon using time series Band 11 (SWIR) of multispectral Sentinel-2 satellite images and machine learning algorithms
No full textSoil Organic Carbon (SOC) is a critical soil property impacting food security and climate change. Traditional methods for SOC estimation are time-consuming, expensive, and unsuitable for large-scale application. Consequently, researchers have increasingly focused on utilizing Remote Sensing (RS) images for SOC estimation over the past two decades. However, achieving high SOC estimation accuracy (more than 80 %) remains challenging. This limitation often stems from a mismatch between the complexity of SOC and the information captured by traditional RS observations (e.g., reflectance bands or spectral indices), as conventional feature extraction methods from RS images may not be detailed enough to monitor the many factors influencing SOC concentration. One promising solution to enhance feature extraction is the use of time series observations, analyzing multiple images over time instead of relying on single-time images. This study proposes a novel approach leveraging time series of the Sentinel-2 satellite's B11 band (centered around 1610 nm, a region sensitive to SOC absorption features) along with Principal Component Analysis (PCA) and Independent Component Analysis (ICA) transformations to extract more meaningful temporal features. Specifically, ten new features based on temporal variations were derived by applying PCA and ICA to the B11 band time series images. These temporal features were then combined with features derived from the median of all Sentinel-2 images acquired during the summer of 2019, corresponding to the soil data collection period. Four machine learning algorithms (RF, GBRT, XGBoost, and LightGBM) were employed across four distinct scenarios to evaluate the novel feature extraction method and a feature selection algorithm. The scenarios were designed as follows: Scenario one (S#1) and Scenario two (S#2) did not utilize the time series features, while Scenario three (S#3) and Scenario four (S#4) did. A binary Genetic Algorithm (GA) for feature selection was implemented in S#2 and S#4, distinguishing them from S#1 and S#3 respectively. XGBoost performed best, achieving an R2 of 0.891 in S#4 (time series features and GA). Incorporating time series features significantly improved accuracy by 0.11, while GA-based feature selection added another 0.05. The findings highlight the effectiveness of the developed feature extraction algorithm, using Sentinel-2's B11 time series and advanced transformations, for substantially improving SOC level estimation
Not Only Text: Exploring Compositionality of Visual Representations in Vision-Language Models
No full textVision-Language Models (VLMs) learn a shared feature space for text and images, enabling the comparison of inputs of different modalities. While prior works demonstrated that VLMs organize natural language representations into regular structures encoding composite meanings, it remains unclear if compositional patterns also emerge in the visual embedding space. In this work, we investigate compositionality in the image domain, where the analysis of compositional properties is challenged by noise and sparsity of visual data. We address these problems and propose a framework, called Geodesically Decomposable Embeddings (GDE), that approximates image representations with geometry-aware compositional structures in the latent space. We demonstrate that visual embeddings of pre-trained VLMs exhibit a compositional arrangement, and evaluate the effectiveness of this property in the tasks of compositional classification and group robustness. GDE achieves stronger performance in compositional classification compared to its counterpart method that assumes linear geometry of the latent space. Notably, it is particularly effective for group robustness, where we achieve higher results than task-specific solutions. Our results indicate that VLMs can automatically develop a human-like form of compositional reasoning in the visual domain, making their underlying processes more interpretable. Code is available at https://github.com/BerasiDavide/vlm_image_compositionality
Three-Phase Multilevel DC/AC Converter and Synergetic Modulation for Split Batteries
No full textThis letter introduces a three-phase multilevel converter for the integration of multiple battery submodules. The circuit comprises a synergetic modulated quasi-single stage design that includes a three-phase three-level voltage source converter operating at low switching frequency and two modular series-connected half-bridge converters operating with high switching frequency. Therein, all DC-link voltage rated devices switch at low frequency and zero voltage while sinusoidal currents are ensured without a complicated control. Furthermore, it provides a multilevel conversion and consequently smaller voltage transients enhancing power quality. Additionally, the modular configuration enables flexible management of the series-connected battery submodules, eliminating the need for an extra balancer between the battery submodules. This paper provides an explanation of circuit operation and the synergistic modulation technique. Simulations and experimental results are presented to validate the feasibility of the proposed circuit