Multimedia ONline ARchiv CHemnitz
Not a member yet
4978 research outputs found
Sort by
Road Construction Site Detection using Low-Level Sensor Fusion for Self-Driving Cars
No full textNavigating through road work zones remains a challenge for the development of
autonomous driving technology. While HD maps are essential for accurate local-
ization and navigation in autonomous vehicles, they face issues when encountering
dynamic and constantly changing situations on the road such as road construction
sites. As a result, autonomous vehicles need to rely on their onboard sensor data
for safe navigation through construction zones. This thesis focuses on low-level
fusion-based methods, using both point cloud and image data for the detection
of road construction sites. The primary objective is to identify temporary traffic
control devices like delineator posts, safety barriers, and traffic cones which play
a role in ensuring road safety while maintaining smooth traffic flow throughout
construction areas.
To achieve this, the CARLA simulator is used to generate an autonomous driving
dataset that represents various road construction sites that are frequently observed
in German regions. This dataset forms the basis for evaluating four state-of-the-
art low-level LiDAR-camera fusion-based methods. By establishing a benchmark,
this thesis presents a proof of concept for successful road work zone detection.
The results demonstrate the effectiveness of low-level fusion-based methods in
identifying road construction sites and open the door for further developments,
emphasizing its potential impact on advancing autonomous driving technology
within work zones
Molecular recognition through copper(I)/NHC bifunctional catalysts enabling site-selective amide hydrogenation
No full textDiese Dissertation behandelt die Reaktivität nukleophiler Kupfer(I)-Hydrid-Komplexe bei der Reduktion von Carbonsäurederivaten unter Verwendung von molekularem Wasserstoff (H2) als terminales Reduktionsmittel. Darüber hinaus wird die Synthese neuartiger bifunktionaler Ligand-Vorstufen diskutiert, die N-heterocyclische Carbene (NHCs) mit verschiedenen organokatalytischen Gruppen innerhalb eines einzelnen Gerüstes kombinieren. Der erste Abschnitt der Arbeit konzentriert sich auf die Synthese verschiedener substituierter Ester für die Kupfer(I)-katalysierte Esterreduktion mit H2.
Aufbauend auf diesen Ergebnissen wird im zweiten Abschnitt eine neuartige Kupfer(I)-katalysierte Methode zur Reduktion von Amiden vorgestellt, bei der ein bifunktionaler Katalysator sowie H2 als Reduktionsmittel eingesetzt werden. Durch die Kombination eines Kupfer(I)/NHC-Komplexes mit einem Guanidin-Organokatalysator reagieren die „weichen“ nukleophilen Kupfer(I)-Hydrid-Komplexe mit „harten“ Carbonsäureamid-Elektrophilen, was zu einer selektiven Reduktion dieser Amide zu den entsprechenden Alkoholen führt. Zudem wurde eine Gruppe hochreaktiver „privilegierter Amide“ identifiziert, wodurch erstmals eine ortsselektive Diamid-Reduktion ermöglicht wurde. Eine umfassende Untersuchung des Substratspektrums sowie mechanistische Studien werden vorgestellt.
Der letzte Abschnitt der Dissertation behandelt die Synthese neuartiger NHC-Ligand-Vorstufen, die Harnstoffe, Thioharnstoffe und Squaramide als organokatalytische Motive enthalten. Dieser Ansatz ermöglicht eine systematische Strategie zur Erzeugung einer vielfältigen Bibliothek von Ligand-Vorstufen. Ureabasierten Kupfer(I)- und Palladium(II)-Komplexe wurden synthetisiert und ihre katalytische Aktivität untersucht
Energy Analysis of Row-Stationary Dataflow in CNN Computation on Microcontrollers
No full textIn this thesis, the Row Stationary (RS) dataflow introduced in Eyeriss is suggested for the MCU. RS maximizes the input reuse and minimizes the partial sum movement. This reduces the energy consumption of the model. The LeNet model is chosen to employ different data types reuse of RS dataflow onto the MCU. The convolutional layers are implemented with input pixel (row-wise) and filter weight reuse (row-wise) and fully connected layers are implemented with input pixel and psum reuse. An analysis framework is considered to compare the conventional and RS dataflow under the same memory area. The CONV layers of RS dataflow save energy and latency up to 14.42 % and 14.42 % respectively in comparison to the conventional dataflow. On the other hand, FC layers save both energy and latency up to 16 %. The energy and latency of the LeNet model are improved by almost 14 % by reducing the SRAM memory accesses for RS dataflow
Software-definierte IP-Optische Netze: Planung, Rekonfiguration und techno-ökonomische Optimierung
No full textSoftware defined networking (SDN) enables the design of multilayer and multi-domain control functions for optimum service and network resource management. These functions require the design of algorithms that exploit the automation capabilities offered by cross-layer and cross-domain SDN control. This thesis investigates the design of these algorithms for networks that integrate packet-switched electrical systems with circuit-switched optical systems. In particular, algorithms are investigated for the planning and online reoptimization of IP-Optical networks. These algorithms consider the resource dependencies between the IP and optical layers in order to perform optimum routing, traffic restoration, network reconfiguration and capacity planning. The algorithms aim at designing networks that supply low-cost services while meeting their quality of service (QoS) requirements. Algorithms are investigated for use cases that include Metro/Core transport, intra-data centre networks and fixed broadband access. Concerning Metro/Core transport, the problem of resilient network design is investigated by applying adjustable robust optimization for the planning of networks that integrate Fast-Reroute with IP-Optical restoration. By this strategy, IP traffic is efficiently restored from optical failures. This is achieved by a linear program that optimizes the network capacity along with the routing and capacity of IP tunnels in failure-free mode and in a selected set of optical failures. Compared to existing traffic restoration approaches, the proposed method significantly reduces the network capacity while providing robustness against unforeseen failures. This problem is also studied from a meta-learning inspired perspective, where an algorithm selection (AS) function learns the relationship between the characteristics of the problem instance (i.e. the size and distribution of the traffic load and an observed optical failure) and the performance of candidate algorithms for traffic restoration. The function provides a hyperheuristic method that predicts the algorithm that requires less capacity to restore traffic for an observed instance. Heuristic methods that integrate IP-Optical restoration are formulated to test the performance of the approach. The results show that the AS function predicts with high accuracy the algorithm that performs best for an observed instance, thereby outperforming existing pure IP-based traffic restoration schemes. The problem of network reoptimization is investigated for transport networks that suffer from spectrum fragmentation in the optical layer. Linear programming is used to formulate algorithms that optimize two objectives, i.e. the spectrum defragmentation state and the strategy whereby optical connections migrate towards their optimized spectrum. The migration strategy minimizes service disruption periods experienced in the IP layer. The proposed algorithms significantly outperform existing defragmentation strategies in both optimization objectives. Regarding intra-data centre networks, a planning approach is proposed for the dimensioning of the network infrastructure and the calculation of the cost of ownership. Compared to networks based on packet-switching, the approach shows that hybrid networks that integrate IP-Optical switching perform well for large data centres. Hybrid networks require less infrastructure and save costs due to reduced energy consumption and repair effort in case of failures. A method for service-oriented optimization is also formulated for the minimization of the direct, shared and common costs incurred for the provisioning of services. Use cases for Metro/Core transport and fixed broadband access show that the method optimizes the economies of scale and scope that arises from the operation of multi-service networks
Verwertung von Abwärme und Sauerstoff zur Steigerung der Rentabilität von Elektrolyseuren in kostenoptimalen Energiesystemen
No full textGrüner Wasserstoff ist ein zentrales Element für die Energie- und Wärmewende, jedoch ist der Betrieb von Elektrolyseuren noch nicht rentabel. Etwa 20 – 50 % der zugeführten Energie gehen während der Elektrolyse verloren, wovon ein Großteil in Form von Wärme und Sauerstoff weitergenutzt werden könnte. Wir untersuchen in einem Optimierungsmodell, wie die Verwertung der Elektrolysenebenprodukte zur Rentabilitätssteigerung beitragen kann, wenn Unsicherheiten über eine Monte-Carlo-Simulation hinreichend berücksichtigt werden. Vier Szenarien veranschaulichen die kostenoptimale Investitions- und Einsatzplanung mit und ohne Allokation der Investitionskosten auf die Elektrolysenebenprodukte. Es zeigt sich, dass Elektrolyseure bei ausgewählten Kombinationen aus Elektrolyseleistungen und Wasserstoffexportpreisen rentabel sein können. Der ROI weist die Rentabilitätssteigerung durch Kostenallokation auf die Elektrolysenebenprodukte nach. Es wird abschließend diskutiert, wie Modell und Methodik weiterentwickelt werden könnten, um die Investitions- und Einsatzplanung unter Unsicherheit für wasserstoffbasierte Energiesysteme zu optimieren.Green hydrogen is a key element of the energy and heat transition, but the operation of electrolyzers is not yet profitable. Around 20 - 50 % of the energy supplied is lost during electrolysis, much of which could be reused in the form of heat and oxygen. In an optimization model, we are investigating how the utilization of electrolysis by-products can contribute to increasing profitability if uncertainties are adequately taken into account using a Monte Carlo simulation. Four scenarios illustrate the cost-optimized investment and deployment planning with and without allocation of investment costs to the electrolysis by-products. It is shown that electrolyzers can be profitable with selected combinations of electrolysis capacities and hydrogen export prices. The ROI demonstrates the increase in profitability through cost allocation to the electrolysis by-products. The paper concludes by discussing how the model and methodology could be further developed to optimize investment and deployment planning under uncertainty for hydrogen-based energy systems
Erstellung eines mathematischen Modells eines brennstoffzellenbetriebenen Mikro-KWK-Systems
No full textIn dieser Arbeit wird ein mathematisches Modell für ein in Haushalten einsetzbares Kraft-Wärme-Kopplungssystem (KWK) auf Basis von PEM-Brennstoffzellen vorgestellt. Unter Anwendung der VDI 4655 wurden die jährlichen Wärme- und Strombedarfskurven für Einfamilienhäuser erstellt. Eine Analyse verschiedener Kapazitäten der Systemkomponenten (Batterie, Wärmespeicher) wurde durchgeführt, um den Einfluss auf den Wasserstoffverbrauch zu untersuchen und die simple payback time (SPT) zu bewerten. Diese Untersuchung liefert wichtige Erkenntnisse für weitere Optimierungsstrategien und ein vertieftes Verständnis sowie die Optimierung von KWK.:1 Einleitung
2 Methodik
3 Ergebnisse
4 Diskussion
5 Zusammenfassung
Literaturverzeichnis
Affiliatio
Design of an Embedded System for Fast Classification of Bimetallic Coins by Impedance Spectroscopy
No full textA novel embedded impedance measurement system for counterfeit detection of bimetallic coins is presented. The system integrates Eddy current sensors with inductive spectroscopy, facilitating the detection of hidden inlay security features and the identification of surface minting. A salient feature of the system is the use of a single Eddy current sensor that operates across multiple frequencies. This feature eliminates the need for extensive calibration procedures typically required when multiple sensor coils are used for each excitation signal frequency. The system employs an undersampling technique, facilitating impedance measurements over a wide frequency in MHz range using a simple microcontroller. The employment of machine learning-based classification further enhances the system's accuracy, enabling precise coin classification.
The system's high-speed classification capability of 203 coins per second leads to a substantial enhancement in counterfeit detection while reducing the system footprint, cost, and power consumption. The classification algorithm has been rigorously tested on multiple datasets with varying difficulty levels, ensuring its robustness and reliability under different conditions. The compact, real-time, and cost-effective design of a system represents a significant breakthrough in modern coin counterfeit detection, setting new standards for accuracy and efficiency.:1. Introduction
2. Theoretical Background
3. State of the Art
4. Bimetallic Coin Investigation
5. Embedded Bimetallic Coin Classification System
6. Conclusion and Outloo
BLUME II Vignette Study: Technical report on the DFG-funded BLUME study on the beliefs of primary school teachers on dealing with the topic of multilingualism
No full textThe BLUME study is designed as a mixed-methods study in which the beliefs of primary school
teachers on dealing with multilingualism are analysed using different methodological approaches.
This technical report refers to the data collection of the DFG-funded BLUME II vignette study
(04/22-3/25), in which an intensive theoretical and qualitative-empirical analysis of teachers' beliefs
about dealing with multilingualism in primary school lessons was carried out. The core of the
BLUME II vignette study is the differentiated analysis of the subject of primary school teachers'
beliefs. In particular, the aim is to explore the complexity and contradictions of the beliefs and to
reveal possible hierarchies within the teachers' belief systems. This technical report presents the
theory-based teaching vignettes developed and describes the piloting phase, the sampling for the
main data collection and the sample composition.:0 The BLUME study: Brief presentation of the sub-studies
1 Teaching vignettes as stimulus in the interview situation
2 Piloting phase
3 Sampling for the main survey
4 Sample description
5 Transcription rules for BLUME interview
Parametrische Baugruppenmethode zur Bestimmung der Denavit-Hartenberg Transformation
No full textDie Bestimmung der Denavit-Hartenberg-Parameter ist oft fehlerhaft und nicht leicht
aufzustellen. In dem Beitrag wird eine parametrische Baugruppenmethode mit einem CAD-System
(Creo Parametrics) vorgestellt. Mit dieser Methode ist es möglich für beliebige Mehrkörpersysteme
wie Industrieroboter eine fehlerfreie DH Tabelle aufzustellen.
Alle Parameter werden in einer zugeordneten Baugruppe bestimmt und können für die weiteren
Transformationen oder inversen kinematischen Betrachtungen verwendet werden. Am Beispiel
eines Knickarmroboters mit sechs Gelenken wird die Methode angewendet und so kann die Sollpose
mit berechneten Gelenkwinkeln erreicht werden.The determination of the Denavit-Hartenberg parameters is often incorrect and not easy to
determine. The article presents a parametric assembly method using a CAD system (Creo
Parametrics). This method makes it possible to set up an error-free DH table for any multi-body
system such as industrial robots.
All parameters are determined in an assigned assembly and can be used for further
transformations or inverse kinematic considerations. The method is applied using the example of
an articulated arm robot with six joints, and the target pose can be achieved with calculated joint
angles
AI integration in simulation processes: practical examples
No full textIf a learning base of different parameter combinations and the CAE results calculated with them is available, AI / machine learning can be used to predict the results for new parameter combinations. The entire CAE model is replaced by the machine learning model (ROM, Reduced Order Model).
This presentation focuses on applications where only part of the CAE model is replaced by a ROM. Typically a computationally intensive part is replaced. Also, if a part of the model should not be passed on for confidentiality reasons, one solution is to replace it with a ROM.
The presentation uses practical examples with the machine learning software Odyssey to show how the interface between ROM and CAE model works. MSC Nastran Smart Superelements and the FMU format are examined in more detail.Wenn eine Learning Base aus verschiedenen Parameterkombinationen und den damit berechneten CAE-Ergebnissen vorhanden ist, können mit KI / Machine Learning die Ergebnisse für neue Parameterkombinationen vorhergesagt werden. Dabei wird das gesamte CAE-Modell durch das Machine Learning Modell (ROM, Reduced Order Model) ersetzt.
In diesem Vortrag wird der Fall betrachtet, dass nur ein Teil des CAE-Modells durch ein ROM ersetzt wird, typischerweise ein rechenintensiver Teil. Auch wenn ein Modellteil aus Geheimhaltungsgründen nicht weitergegeben werden soll, ist eine Lösung, ihn durch ein ROM zu ersetzen.
Der Vortrag zeigt an praktischen Beispielen mit der Machine Learning Software Odyssee, wie das Interface zwischen ROM und CAE-Modell funktioniert. MSC Nastran Smart Superelements und das FMU-Format werden genauer betrachtet