Repository der Technischen Hochschule Ingolstadt
Not a member yet
4170 research outputs found
Sort by
Dynamische Optimierung von VR-Lernumgebungen durch generative KI
No full textDie steigende Leistungsfähigkeit von Large Language Models (LLMs) eröffnet neue Wege für die automatisierte Erstellung virtueller Lernwelten. In diesem Beitrag stellen wir einen Ansatz vor, mit dem VR-Szenarien allein durch natürlichsprachliche Beschreibungen generiert und dynamisch an Lerninhalte angepasst werden können. Dabei setzt unser System auf Diffusionsmodelle und eine metasprachliche Strukturierung (MLDS), um Objekte und didaktische Elemente bedarfsgenau und konsistent anzuordnen. Erste Prototypen zeigen, dass sich dadurch VR-Lernumgebungen effizienter und flexibler erstellen lassen, was insbesondere für MINT-Fächer wertvolle Potenziale bietet. Lehrende können so Lernlabore virtuell nachbilden und im laufenden Kurs an geänderte Anforderungen anpassen, ohne umfangreiche 3D-Design-Kenntnisse zu benötigen
Enhancing Indoor Localization Accuracy based RSSI Fingerprint with CNN, Modified VGG16 and Modified AlexNet
No full text
Clustering Dutch citizens into behavioural phenotypes to understand green energy investment preferences
No full textPeople differ in their underlying economic preferences and needs for energy retrofits. Accelerating the energy transition, therefore, requires tailoring personalised solutions for distinct groups of individuals. In this paper, we create behavioural phenotypes of green energy investors in the residential sector of the Netherlands. Using a latent class analysis on a representative sample of 2245 respondents, we identify four distinct classes of investors: Comfort-driven Rationalists, Financially Driven Rationalists, Policy-driven Environmentalists, and Erratic Choosers. We innovate upon the literature by linking class profiling to economic preferences and behavioural biases, alongside socio-demographic and household characteristics. Our findings can help practitioners design bottom-up tailored behavioural interventions to accelerate the uptake of green energy investments
CoDa-4DGS: Dynamic Gaussian Splatting with Context and Deformation Awareness for Autonomous Driving
Get PDFDynamic scene rendering opens new avenues in autonomous driving by enabling closed-loop simulations with photorealistic data, which is crucial for validating end-to-end algorithms. However, the complex and highly dynamic nature of traffic environments presents significant challenges in accurately rendering these scenes. In this paper, we introduce a novel 4D Gaussian Splatting (4DGS) approach, which incorporates context and temporal deformation awareness to improve dynamic scene rendering. Specifically, we employ a 2D semantic segmentation foundation model to self-supervise the 4D semantic features of Gaussians, ensuring meaningful contextual embedding. Simultaneously, we track the temporal deformation of each Gaussian across adjacent frames. By aggregating and encoding both semantic and temporal deformation features, each Gaussian is equipped with cues for potential deformation compensation within 3D space, facilitating a more precise representation of dynamic scenes. Experimental results show that our method improves 4DGS's ability to capture fine details in dynamic scene rendering for autonomous driving and outperforms other self-supervised methods in 4D reconstruction and novel view synthesis. Furthermore, CoDa-4DGS deforms semantic features with each Gaussian, enabling broader applications
Influence of Blind Spot Assistance Systems in Heavy Commercial Vehicles on Accident Reconstruction
Get PDFAccidents between right-turning commercial vehicles and crossing vulnerable road users (VRU) in urban environments often lead to serious or fatal injuries and therefore play a significant role in forensic accident analysis. To reduce the risk of accidents, blind spot assistance systems have been installed in commercial vehicles for several years, among other things, to detect VRUs and warn the driver in time. However, since such systems cannot reliably prevent all turning accidents, an investigation by experts must clarify how the accident occurred and to what extent the blind spot assistance system influenced the course of the accident. The occurrence of the acoustic warning message can be defined as an objective reaction prompt for the driver, so that the blind spot assistance system can significantly influence the avoidability assessment. In order to be able to integrate the system into forensic accident analysis, a precise knowledge of how the system works and its limitations is required. For this purpose, tests with different systems and accident constellations were conducted and evaluated. It was found that the type of sensor used for the assistance systems has a great influence on the system’s performance. The lateral distance between the right side of the commercial vehicle and the VRU as well as obstacles between them and the speed difference can take great influence on the reliability of the assistance system. Depending on the concrete time of the system’s warning signal the accident can be avoided or not by the driver when reacting on this signal
Additive manufacturing of a bridge in situ
Get PDFAdditive manufacturing (AM) is playing an increasing role in the production of complex steel structures. Robot‐ or machine‐guided gas‐shielded metal arc welding (GMAW), known as wire and arc additive manufacturing (WAAM), is suitable for this purpose. A small footbridge was designed in shell form at TU Darmstadt. It was completely additively manufactured over a stream on site using a welding robot. The work called for cantilevered manufacturing without support structures, which poses special challenges in manufacturing. This paper describes selected aspects of this project: the preliminary strength studies, the manufacturing strategy to ensure homogeneous manufacturing and the findings from on‐site manufacturing. In addition, investigations were carried out which led to a notable increase in the deposition rate with the right choice of gas and wire
Uncertainty-Aware Hybrid Machine Learning in Virtual Sensors for Vehicle Sideslip Angle Estimation
Get PDFPrecise vehicle state estimation is crucial for safe and reliable autonomous driving. The number of measurable states and their precision offered by the onboard vehicle sensor system are often constrained by cost. For instance, measuring critical quantities such as the Vehicle Sideslip Angle (VSA) poses significant commercial challenges using current optical sensors. This paper addresses these limitations by focusing on the development of high-performance virtual sensors to enhance vehicle state estimation for active safety. The proposed Uncertainty-Aware Hybrid Learning (UAHL) architecture integrates a machine learning model with vehicle motion models to estimate VSA directly from onboard sensor data. A key aspect of the UAHL architecture is its focus on uncertainty quantification for individual model estimates and hybrid fusion. These mechanisms enable the dynamic weighting of uncertainty-aware predictions from machine learning and vehicle motion models to produce accurate and reliable hybrid VSA estimates. This work also presents a novel dataset named Real-world Vehicle State Estimation Dataset (ReV-StED), comprising synchronized measurements from advanced vehicle dynamic sensors. The experimental results demonstrate the superior performance of the proposed method for VSA estimation, highlighting UAHL as a promising architecture for advancing virtual sensors and enhancing active safety in autonomous vehicles