KITopen
Not a member yet
63711 research outputs found
Sort by
Parameter Tuning Under Uncertain Road Perception in Driver Assistance Systems
Full text linkAdvanced driver assistance systems have improved comfort, safety, and efficiency of modern vehicles. However, sensor limitations lead to noisy lane estimates that pose a significant challenge in developing performant control architectures. Lateral trajectory planning often employs an optimal control formulation to maintain lane position and minimize steering effort. The parameters are often tuned manually, which is a time-intensive procedure. This paper presents an automatic parameter tuning method for lateral planning in lane-keeping scenarios based on recorded data, while taking into account noisy road estimates. By simulating the lateral vehicle behavior along a reference curve, our approach efficiently optimizes planner parameters for automated driving and demonstrates improved performance on previously unseen test data
Predicting Air Flow in Calendered Paper Sheets from μ-CT Data: Combining Physics with Morphology
Full text linkPredicting the macroscopic properties of thin fiber-based porous materials from their microscopic morphology remains challenging because of the structural heterogeneity of these materials. In this study, computational fluid dynamics simulations were performed to compute volume air flow based on tomographic image data of uncompressed and compressed paper sheets. To reduce computational demands, a pore network model was employed, allowing volume air flow to be approximated with less computational effort.
To improve prediction accuracy, geometric descriptors of the pore space, such as porosity, surface area, median pore radius, and geodesic tortuosity, were combined with predictions of the pore network model. This integrated approach significantly improves the predictive power of the pore network model and indicates which aspects of the pore space morphology are not accurately represented within the pore network model. In particular, we illustrate that a high correlation among descriptors does not necessarily imply redundancy in a combined prediction
The annual variation of the M2 gravimetric tidal parameters investigated with nonlinear, time-stepping ocean models
Full text linkTemporal variations of the M2 tidal parameters in gravity are observed at all superconducting gravimeter stations. We specifically investigate the annual variation of M2 tidal parameters. A similar variation is observed for the parameters from sea surface heights which is larger than expected from astronomical forcing alone. This leads to the hypothesis that the variations of the gravimetric tidal parameters are caused by the loading of the annual variation of M2 in the oceans. Only nonlinear, time-stepping ocean models are able to describe such variations. We use sea surface heights from three global and two regional models of this kind to calculate the loading. The loading time series is then added to synthetic body tides and analyzed by a moving window tidal analysis with ETERNA in the same way as the measured data. We compare the resulting variations of the M2 tidal parameters for synthetic gravity with those observed from measurements. Three of the five ocean models show an annual variation of a similar order of magnitude which supports our hypothesis. The other two ocean models produce smaller or no clear annual variation of the M2 tidal parameters. In the ocean the annual variation of M2 has large amplitudes in shelf areas and small amplitudes in the open ocean. Large areas with small amplitude might contribute to the gravity loading as much as small areas with large amplitudes do. We investigate this with the global Hycom model at three SG stations. The investigation shows that not only close shelf areas but also distant ocean regions, including open ocean areas, contribute significantly to the annual variation of the M2 tidal parameters at the superconducting gravimeter stations
Cross-Validation of Muon Content in Extensive Air Showers with Surface and Underground Detectors of AugerPrime
Full text linkNonlinear dynamics of periodic Lugiato-Lefever waves against sums of co-periodic and localized perturbations
Full text linkIn recent years, essential progress has been made in the nonlinear stability analysis of periodic Lugiato-Lefever waves against co-periodic and localized perturbations. Inspired by considerations from fiber optics, we introduce a novel iteration scheme which allows to perturb against sums of co-periodic and localized functions. This unifies previous stability theories in a natural manner
Residual stress in Germanium single crystals caused by femtosecond laser micromachining
Full text linkFemtosecond laser (fs-laser) milling has emerged as a promising technique for high-precision material processing, offering significantly faster ablation rates compared to Ga+ Focused Ion Beam (FIB) milling. While fs-laser ablation is often considered to be athermal, its impact on surface features, such as redeposited material, raises concerns about its influence on microstructure and residual stress fields. This study explores the mechanical effects of fs-laser and FIB milling on a germanium single crystal, using synchrotron-based Laue microdiffraction coupled with Digital Image Correlation to characterize induced residual stresses and their spatial distribution. The further development of this technique allows to push the strain resolution to 10⁻⁵, which enabled a clear identification of the influence of the redeposition structure
Spatiotemporal scenarios of socioeconomic futures in Germany
Full text linkSocioeconomic development influences both the drivers and consequences of climate change, but many scenario applications still rely on highly aggregated indicators such as GDP and population, which mask regional diversity. This study develops spatially explicit socioeconomic scenarios for Germany to support climate action and land-use planning with greater detail and contextual relevance. Using a mixed-methods framework, we integrate historical trend analysis, participatory scenario building, and quantitative projection to generate annual trajectories of key indicators at district level from 2020 to 2100. The indicators cover human, social, financial, and manufactured capital, including demographic dynamics, education, income, employment, inequality, and social cohesion. We analyse the dataset with correlation and clustering methods to explore interdependencies and to identify distinct regional development pathways. Results highlight persistent associations between income, education, and life expectancy, but also scenario-specific changes in the relations between inequality, employment, and urbanisation. Strong east–west disparities and urban–rural contrasts remain across all scenarios, while a sufficiency-oriented pathway demonstrates that wellbeing gains can occur without economic growth. By providing high-resolution, multidimensional socioeconomic scenarios, this study enhances integrated climate–land modelling and informs the design of regionally adaptive and socially equitable climate policies under multiple plausible futures
Scaffolding Dexterous Manipulation with Vision-Language Models
Full text linkDexterous robotic hands are essential for performing complex manipulation tasks, yet remain difficult to train due to the challenges of demonstration collection and high-dimensional control. While reinforcement learning (RL) can alleviate the data bottleneck by generating experience in simulation, it typically relies on carefully designed, task-specific reward functions, which hinder scalability and generalization. Thus, contemporary works in dexterous manipulation have often bootstrapped from reference trajectories. These trajectories specify target hand poses that guide the exploration of RL policies and object poses that enable dense, task-agnostic rewards. However, sourcing suitable trajectories - particularly for dexterous hands - remains a significant challenge. Yet, the precise details in explicit reference trajectories are often unnecessary, as RL ultimately refines the motion. Our key insight is that modern vision-language models (VLMs) already encode the commonsense spatial and semantic knowledge needed to specify tasks and guide exploration effectively. Given a task description (e.g., "open the cabinet") and a visual scene, our method uses an off-the-shelf VLM to first identify task-relevant keypoints (e.g., handles, buttons) and then synthesize 3D trajectories for hand motion and object motion. Subsequently, we train a low-level residual RL policy in simulation to track these coarse trajectories or "scaffolds" with high fidelity. Across a number of simulated tasks involving articulated objects and semantic understanding, we demonstrate that our method is able to learn robust dexterous manipulation policies. Moreover, we showcase that our method transfers to real-world robotic hands without any human demonstrations or handcrafted rewards
Scaffolding Dexterous Manipulation with Vision-Language Models
Full text linkDexterous robotic hands are essential for performing complex manipulation tasks, yet remain difficult to train due to the challenges of demonstration collection and high-dimensional control. Thus, contemporary works in dexterous manipulation have often bootstrapped from reference trajectories to trajectories specify target hand poses that guide the exploration of RL policies and object poses that enable dense, task-agnostic rewards. However, sourcing suitable trajectories---particularly for dexterous hands---remains a significant challenge. Our key insight is that modern vision-language models (VLMs) already encode the commonsense spatial and semantic knowledge needed to specify tasks and guide exploration effectively. Given a task description (e.g., “open the cabinet”) and a visual scene, our method uses an off-the-shelf VLM to first identify task-relevant keypoints (e.g., handles, buttons) and then synthesize 3D trajectories for hand motion and object motion. Subsequently, we train a low-level residual RL policy in simulation to track these coarse trajectories or ``scaffolds\u27\u27 with high fidelity