Hochschule Konstanz University of Applied Sciences
Hochschulschriftenserver der HTWG KonstanzNot a member yet
2319 research outputs found
Sort by
Hybrid Bernstein Normalizing Flows for Flexible Multivariate Density Regression with Interpretable Marginals
No full textDensity regression models allow a comprehensive understanding of data by modeling the complete conditional probability distribution. While flexible estimation approaches such as normalizing flows (NFs) work particularly well in multiple dimensions, interpreting the input-output relationship of such models is often difficult, due to the blackbox character of deep learning models. In contrast, existing statistical methods for multivariate outcomes such as multivariate conditional transformation models (MCTMs) are restricted in flexibility and are often not expressive enough to represent complex multivariate probability distributions. In this paper, we combine MCTMs with state-of-the-art and autoregressive NFs to leverage the transparency of MCTMs for modeling interpretable feature effects on the marginal distributions in the first step and the flexibility of neuralnetwork-based NF techniques to account for complex and non-linear relationships in the joint data distribution. We demonstrate our method’s versatility in various numerical experiments and compare it with MCTMs and other NF models on both simulated and real-world data
Trajectory Planning with Signal Temporal Logic Costs Using Deterministic Path Integral Optimization
No full textFormulating the intended behavior of a dynamic system can be challenging. Signal temporal logic (STL) is frequently used for this purpose due to its suitability in formalizing comprehensible, modular, and versatile spatiotemporal specifications. Due to scaling issues with respect to the complexity of the specifications and the potential occurrence of non-differentiable terms, classical optimization methods often solve STL-based problems inefficiently. Smoothing and approximation techniques can alleviate these issues but require changing the optimization problem. This paper proposes a novel sampling-based method based on model predictive path integral control to solve optimal control problems with STL cost functions. We demonstrate the effectiveness of our method on benchmark motion planning problems and compare its performance with state-of-the-art methods. The results show that our method efficiently solves optimal control problems with STL costs.
©2025 IEE
Millisecond NMPC for Swing-Up and Stabilization of the Furuta Pendulum in Real World
No full textThe Furuta pendulum’s swing-up and stabilization control is currently used by many researchers to benchmark nonlinear control algorithms. In this brief, we give a systematic overview of important contributions to the control of the Furuta pendulum presented in the last 15 years. Furthermore, we use nonlinear model predictive control (NMPC) to design a real-time capable holistic controller. An optimal control problem (OCP) including a detailed nonlinear system dynamics model is defined, transcribed into a nonlinear optimization problem via direct multiple shooting, and solved in real time on an embedded system using acados. A breakthrough concerning the control performance was achieved by the usage of efficient discretization via a nonuniform grid, solving the tradeoff between a long prediction horizon and a fast sample time. The control strategy shows excellent performance in simulation and real-world experiments using a custom-made pendulum prototype
E-Assessment am Beispiel eines Auswahlverfahrens für internationale Studierende
No full textDie Internationalisierung von Hochschulen sowie das wachsende Interesse an
internationalen Studierenden legen nahe, dass neue Wege bei der Rekrutierung
und Auswahl beschritten werden. Insbesondere, wenn deutsche Hochschulen
die Eignung von internationalen Studienbewerbern prüfen möchten, die sich
nicht am Hochschulstandort befinden, bieten sich E-Assessments an. Allerdings
sind im Kontext von Online-Prüfungen technische, organisatorische,
rechtliche und ethische Fragen zu klären. Im vorliegenden Beitrag werden
diese Fragen am Beispiel eines Auswahlverfahrens für internationale Studierende
erläutert
Feasibility and Signal Quality Evaluation of a Sleep Apnoea Detection System Using Wearable Sensor Technology
No full textAdequate and restorative sleep is essential for maintaining physiological health, cognitive function and overall well-being. However, sleep disorders - particularly obstructive sleep apnoea (OSA) - can significantly impair sleep quality and contribute to the development of other complications. Despite its high prevalence, OSA often remains undiagnosed, largely due to the limited accessibility and convenience of current diagnostic procedures. To address this challenge, the present study with 20 participants focuses on the development and preliminary evaluation of a patient-centred system designed for automated detection of sleep apnoea and long-term monitoring of therapeutic interventions. The system integrates electrical impedance pneumography and photoplethysmography to record key physiological parameters including respiratory effort, heart rate and peripheral oxygen saturation (SpO₂). System modules were implemented for signal processing, data analysis and wireless transmission. An initial evaluation of the prototype was conducted to assess signal quality. While the results indicate several areas for optimisation - particularly in terms of hardware stability and reliability of data transmission - the overall concept shows considerable potential for future application in clinical and home environments
Sleep-Driven Haptic Stimulation for Resilient Somatosensory Rehabilitation
No full textTouch, as one of the five primary senses, provides critical somatosensory information, including pressure, vibration, temperature, pain, and skin stress. Somatosensory deficits, often resulting from stroke or aging, significantly impair fingertip sensitivity, affecting daily function and quality of life. Vibrotactile stimulation devices have emerged as a modern therapeutic approach to address these deficits. A well-established bidirectional relationship exists between somatosensory function and sleep, where somatosensory stimulation aids sleep regulation, and sleep enhances somatosensory recovery. However, the potential of somatosensory therapy in ambulatory settings remains largely unexplored. While MRI and EEG have been used to measure the effects of somatosensory stimulation, current therapeutic evaluations still rely primarily on patient feedback, highlighting the need for objective assessment methods. This research initiates an in-depth literature review on the interplay between somatosensory therapy and sleep, alongside the application of EEG for therapy evaluation in home environments. The study aims to guide data selection strategies by sourcing information from patients, therapy sessions, sleep monitoring, and EEG recordings. A system architecture will be designed to integrate somatosensory therapy with sleep monitoring and EEG, addressing hardware requirements, communication protocols, and information architecture. Furthermore, deep learning models will be developed to analyze the interaction between sleep and somatosensory therapy, enabling personalized therapy adaptations
Circular Detection-Driven Adaptive Birth Densities
No full textMultiobject tracking filters require birth densities for the detection of new objects from measurement data. However, modeling these densities is challenging since the expected initial states of newborn objects are rarely known and may change over time. In this paper, we propose circular adaptive birth densities covering up to 360◦ around the sensor. These densities are primarily described in polar coordinates, accounting for point-symmetric quantities such as radius, radial velocity, and tangential velocity of objects entering the surveillance area. It is assumed that these quantities are unknown and may change over time. Therefore, the initial states of successfully detected trajectories are used to estimate the birth density. To accommodate filters operating on Cartesian coordinates, we approximate the adapted birth density using a Gaussian mixture. Furthermore, we consider the influence of ego motion, which impacts both the detection probability and the expected initial states of other objects. Simulation studies show that our approach estimates the birth density parameters based on previously observed objects without bias and that the ego motion consideration improves the precision of the initialization of velocities significantly. The relevance to real-world applications is demonstrated in a maritime scenario
Energiewende in Deutschland
No full textDie Energiewende ist ein Transformationsprozess, der Herausforderungen in unterschiedlichen Themengebieten mit sich bringt, und deshalb ganzheitlich betrachtet werden muss. Dieses Buch bietet einen Einstieg in die komplexe Thematik, indem es die Energiewende in Deutschland aus unterschiedlichen Blickwinkeln, vor allem aus traditioneller industrie- und institutionenökonomischer sowie evolutorisch-systemischer Sicht, kompakt darstellt sowie die vielfältigen Herausforderungen und potenzielle Lösungen aufzeigt. Neben den Ursprüngen der deutschen Energiewende betrachtet es dazu auch die aktuelle Phase mit der Wärmewende und der Sektorkopplung im Mittelpunkt und gibt einen Ausblick auf die nächsten Phasen. Zudem diskutiert es eine mögliche deutsche Vorreiterrolle bei der notwendigen globalen Energiewende.
Für die 2. Auflage wurden alle Inhalte umfassend aktualisiert und überarbeitet sowie neue politische Entwicklungen auf nationaler, europäischer und geopolitischer Ebene - insbesondere der Russland-Ukraine-Konflikt und die „Rush-to-burn“-Problematik - berücksichtigt