05 Fakultät Informatik, Elektrotechnik und Informationstechnik
Permanent URI for this collectionhttps://elib.uni-stuttgart.de/handle/11682/6
Browse
503 results
Search Results
Item Open Access Modelllierung von TOSCA-basierten Deployment Modellen in Java(2020) Fuhrmann, JanIn den letzten Jahren gewinnt das Thema Cloud Computing in vielen Bereichen der Gesellschaft zunehmend an Bedeutung. Immer mehr Unternehmen verlagern Anwendungen sowie Teile der lokalen Infrastruktur in die Cloud. Die Automatisierung des Deployments solcher Cloud-Anwendungen wird jedoch durch deren steigende Komplexität und zugleich hohe Diversität signifikant erschwert, weshalb eine einheitliche Strukturierung erforderlich ist. Die „Topology and Orchestration Specification for Cloud Applications (TOSCA)“ definiert einen Standard für die Beschreibung und das Management sowie die automatisierte Bereitstellung von Cloud-Anwendungen. Alle Komponenten einer Anwendung, in TOSCA Node Templates genannt, inklusive Abhängigkeiten und Verbindungen (Relationship Templates) werden dabei innerhalb sogenannter Topology Templates dargestellt. Bisher erfolgt die Modellierung von TOSCA Topologien ausschließlich in XML oder YAML, aber in einer objektorientierten Programmiersprache wie Java ist dies bislang nicht möglich. In dieser Arbeit wird ein Konzept vorgestellt, das eine automatisierte Konvertierung von TOSCA Node Types und Relationship Types in die Programmiersprache Java beschreibt. Diese Umwandlung findet im sogenannten „TOSCA2JAVA-Transformator“ statt, der im Rahmen dieser Arbeit entwickelt wurde. Er bildet damit die Grundlage für eine zukünftige Kreierung und Konfiguration von TOSCA Topologien in der objektorientierten Umgebung von Java.Item Open Access Determining carbon dioxide footprint using data center simulators(2022) Keppler, MichaelThe modern world relies more and more on highly available software and digital data. To make this possible, the demands on data centers are increasing to provide the services in the required quality. This is associated with an increase in their power consumption and thus also their CO2 emissions. Today, they are already among the largest consumers of electricity and producers of CO2. Due to climate change and the set target of limiting global warming to a maximum of 2° Celsius, the aim is to reduce CO2 emissions. To calculate the CO2 footprint of data centers, data center simulators are already used in their planning phases. Simulators provide operators with important information about the future data center and show room for improvement. This thesis deals with the calculation of the Co2 footprint of data centers, the associated factors and provides an extended approach to the calculation of the CO2 footprint. Using a life cycle assessment approach, it is shown how the carbon footprint of data centers can be comprehensively calculated. For this purpose, several life cycle stages are determined that contribute to the CO2 footprint during the life cycle of a data center. Based on the life cycle stages, requirements for simulators are developed and, with the help of these, four current data center simulators are evaluated in their ability to calculate a CO2 footprint. Since current simulators calculate the CO2 footprint only partially and incompletely, the analysis of this work shows the need for a new simulator. Based on the analyzed life cycle stages, a new simulator is therefore designed. The subsequent evaluation of the simulator implemented as a proof-of-concept against the established requirements shows that the CO2 emissions of a data center can be comprehensively calculated with the demonstrated concept. In addition, a case study is carried out based on the High-Performance Computing Center Stuttgart and compared with the environmental statement of the HLRS.Item Open Access Towards a neuro-symbolic approach for occupant activity recognition : combining temporal HTN planning with hidden Markov models(2025) Hösch, PeterThe problem of occupant activity recognition has gained in relevance due to demographic shifts and growing environmental concerns where context-sensitive applications promise to help. The prevalent approach to this problem is based around the use of supervised machine learning, which faces challenges due to its requirement for large amounts of annotated training data and its tendency to overfit. Using preexisting common sense or expert knowledge, usually in the form of ontologies, presents another option, but carries its own set of shortcomings. Recently, the usage of hierarchical task network planning as an alternative to this ontological approach has been proposed. Hybrid systems that utilize both machine learning and preexisting knowledge promise to preserve the strength of both approaches while alleviating their drawbacks. We propose a new hybrid occupant activity system using hierarchical task network planning to support the training of a Hidden Markov Model, which, to the best of our knowledge, has not been done before. In addition, we evaluate the system on real sensor data in order to find out how much merits this new design has. Hereby we attempt and compare multiple approaches to the problem. Although not all methods improve the performance, the results show that the basic idea is sound and can generate measurable improvements.Item Open Access Service-Based Translation of Quantum Circuits(2022) Kuhn, Maximilian Jakob JohannesRecently, quantum advantage has started to attract attention to the field of quantum computing. While current devices are still noisy and error-prone, numerous vendors have already established themselves, each offering their various approaches with different characteristics and optimizations. In the era of Noisy Intermediate-Scale Quantum (NISQ) computers, quantum circuits must be compiled and executed as efficiently as possible, to best utilize the limited quantum resources available. Therefore, selecting a fitting vendor is a major part of programming for quantum devices. However, different vendors offer different, often incompatible frameworks. Compiled circuits are also highly complex, making manual comparison non-trivial. The NISQ Analyzer has been presented as a solution to this issue. It automates the compilation process of a circuit over a subset of usually incompatible providers. For this purpose it utilizes translation, allowing it to access multiple frameworks even with a circuit only provided in one language. In this thesis, we extend upon this functionality. We make new frameworks available for translation, employing existing translation functionality where possible. For proof of concept, we also implement compilation for a new vendor using the NISQ Analyzer, utilizing our translations. Additionally, we include a detailed evaluation of the reliability of translation frameworks, as well as a case study showing how our extensions can be put to use.Item Open Access Concept for executing management operations on components of application instances(2019) Sowoidnich, YannicA large field of technologies exist for orchestrating cloud applications. Many of them focus on automated deployment techniques, rather than continous management of application instances. Executing operations for deploying applications is different from executing management operations, due to their dependencies to the application state. Proper state management is important to guarantee valid execution of management operations. Cloud providers such as Amazon have embedded functions for managing cloud applications, but they come with major drawbacks. They increase vendor-dependency and they do not support multi-cloud deployments. Technologies like Chef, Puppet or Terraform work with declarative process models, which cannot be used for non-state-changing operations and they mostly only allow simple operations. It is impossible to execute more customized fine grained operations with those technologies. Also, most of these management tools only support executing operations on the whole application, not on specific components of the application. The objective of this thesis is to find a way for executing management operations on running application instances by combining the information of the deployment model with the instance model of the application. The conceptual approach proposed in this thesis will consider and solve above addressed issues, as well as ensuring proper state management of application instances. The practical feasibility of this concept is validated by a prototypical implementation based on the TOSCA standard and the OpenTOSCA ecosystem.Item Open Access Goal-driven context-sensitive production processes : a case study using BPMN(2016) Kar, DebasisThe Fourth Industrial Revolution, also known as Industry 4.0 or Industrial Internet, predicts that Smart Factories driven by Internet of Things (IoT) and Cyber-Physical Systems, will reinvent the traditional manufacturing industry into a digitalized, a context-aware, and an automated manufacturing that will flourish with contemporary Information and Communication Technology (ICT). As the IoT are being deployed across production cites of the manufacturing companies, the need of decision making inside a business process based upon the received contextual data such as employee availability, machine status, etc. from the execution environment has transpired. Production processes need to be updated and optimized frequently to stay competitive in the market. Context-sensitive Adaptive Production Processes is an adept concept that illustrates how a business process can be context-sensitive keeping itself aligned with the abstract organizational goals. The notion of Context-sensitive Adaptive Production Processes leads us to Context-sensitive Execution Step (CES), a logical construct, that encompasses multiple alternative processes, albeit the best-fitting alternative can only be selected, optimized, and executed in runtime. Realization of the context-sensitive business processes requires a model-driven approach. Being Business Process Model and Notation (BPMN) the de-facto standard for business processes modeling, business experts of manufacturing companies can use custom CES construct of BPMN to model and execute context-sensitive business processes in a model-driven approach. This case study is based upon a scenario where there exists multiple alternatives to achieve the same goal in production, nevertheless all the alternatives are not suitable at a certain point of time as changes in business objectives and execution environment makes adaption tougher. Properties of intelligent production processes are different from traditional processes. Such properties along with the scrutinized properties of standard BPMN facilitates modeling CES integrated processes in BPMN. From the requirements inferred from these properties, standard BPMN is extended with extensions such that context-sensitive business processes can be modeled and executed seamlessly. Developed extensions include a new type of process construct and a new type of process definition that are technology agnostic. Thus, CES approach provides a comprehensive solution that makes production processes contextsensitive as well as goal-driven in unison.Item Open Access Quantenunterstütztes Clustering mit hybriden neuronalen Netzen(2021) Wundrack, PhilippMaschinelles Lernen und Quantencomputer sind zwei aktuelle Forschungsthemen, die großes Potenzial haben. Aktuell wird erforscht, wie diese beiden Gebiete kombiniert werden können, um voneinander zu profitieren. In diesen Bereich fällt die vorliegende Arbeit. In dieser Arbeit wird untersucht, ob hybride neuronale Netze genutzt werden können, um die Ergebnisse von Clustering-Algorithmen zu verbessern. Hierzu wird auf den Daten Dimensionsreduktion mit hybriden Autoencodern durchgeführt, bevor die Daten den Clustering-Algorithmen übergeben werden. Als Ergebnis konnte festgestellt werden, dass für bestimmte Datensätze Clustering-Algorithmen bessere Cluster erstellen können, wenn Dimensionsreduktion mit hybriden Autoencodern durchgeführt wurde, anstatt mit klassischen Autoencodern oder PCA.Item Open Access Development of a vendor independent quantum computing transpiler(2020) Wangler, ThomasIn 2019 a quantum computer took 200 seconds for a task for which a state-of-the-art classical supercomputer would need roughly 10,000 years. Quantum supremacy was shown and the development of quantum technology proceeds quickly. However, several aspects limit the implementation and execution of quantum circuits. The vendors of quantum computers provide their own proprietary software development kit (SDK) to execute quantum circuits on their quantum processing units (QPUs). Furthermore, the computers differ regarding various properties like the native gate set and the qubit connectivity. Therefore, quantum circuits cannot be executed on arbitrary QPUs, but there exists a tight coupling between a QPU and a quantum algorithm implementation. To decouple the development of a circuit from the QPU, this thesis proposes a quantum circuit analysis and transpilation framework that integrates quantum SDKs and enables the comparison and analysis of quantum circuits, as well as the export of executable circuits in the respective assembly language of QPUs from different vendors. Currently existing QPUs, also called Noisy Intermediate-Scale Quantum (NISQ) machines, have a very limited number of qubits and are prone to failure. Thus, the integration of different QPUs enhances the possibilities of quantum circuit developers and avoids the SDK lock-in. Additionally, a graphical user interface is developed to support the user in the whole process of importing, visualizing, editing, and simulating a quantum circuit, as well as, choosing a suitable QPU to execute the circuit.Item Open Access Occlusion handling in behavior planning using imitation learning for autonomous driving(2022) Palaniswamy, JanaranjaniCommissioning a self driving vehicle to run on road, requires the facilitation of complete vehicle system to work at all conditions. Behavior planning is a crucial part of the autonomous driving system and it is important to ensure safe and comfortable navigation of the ego vehicle. More advancements are required to enhance the data-driven approaches for the planning systems. The urban driving scenarios always possess a variety of disturbances and inefficiencies. In which, the roundabout is a challenging driving task where uncertainties are caused due to static priority rules and occlusions that limits the field of view for the ego vehicle. Thus behavior planning must make sure to consider the uncertainty of limited visibility of the environment explicitly. Although machine learning-based approaches show promising results for behavior planning. A single planner cannot handle all other urban driving scenarios. Hence, an imitation learning-based technique can help the behavior planner to mimic the human expert behavior. In this context, an end-to-end planning system based on imitation learning proposed by Waymo is used. The behavior planning framework makes use of mid-level input and output representations making it viable to be interfaced with existing vehicle system. The planner outputs a set of waypoints to drive the vehicle controller. However, the existing imitation learning-based planning framework with the Intelligent Driver Model (IDM) as an expert and policy model made of a multi-task network did not address this use case of occluded roundabouts. As the default IDM generates training data with a visibility of the environment, there arises a need for a strategic approach to handle the occluded environments. This thesis work aims at leveraging the existing planning system to handle the situations in a roundabout with limited visibility. Ultimately, the goal is to train the policy model with more realistic data and enable it to make safe and comfortable driving decisions. For this purpose, an occlusion algorithm is implemented to induce limited visibility of the roundabout environment in simulation. And the expert model is enhanced to handle the limited field of view much similar to how a human driver behaves. Consequently, the training dataset generated from the expert is upgraded with an additional input feature. This add-on feature in the input data provides enough knowledge for the policy to perform well in the occluded environment. A study on modern architecture search is performed and a suitable convolutional network is adopted as the backbone for this multi-task model. The enhanced behavior of the proposed approach is demonstrated via detailed quantitative analysis. For this purpose, a new comfort metric is defined and used as Key performance Indicator (KPI) to evaluate the models. An ablation study is conducted with the expert and confirmed that the new extended IDM behaves more carefully in an occlusion environment. In the end, the influence of the training data is inferred by a detailed comparison of the policy model in default and occlusion environments with different dataset configurations. The importance of more realistic data is realized and also shows that the policy model can imitate the expert behavior well enough. It is exhibited that the proposed methodology can handle the occlusions in the complex roundabout situations in simulation.Item Open Access Optimizing the efficiency of data-intensive Data Mashups using Map-Reduce(2017) Sarangi, SunayanaIn order to derive knowledge and information from data through data processing, data integration and data analysis, a variety of Data Mashup tools have been developed in the past. Data Mashups are pipelines that process and integrate data based on different interconnected operators that realize data operations such as filter, join, extraction, alteration or integration. The overall goal is to integrate data from different sources into a single one. Most of these Mashup tools offer a grahical modeling platform, enabling the users to model the data sources, data operations and the data flow, thus, creating a so called Mashup Plan. This enables non-IT experts to perform data operations without having to deal with their technical details. Further, by allowing easy re-modeling and re-execution of the Mashup Plan, it also allows an iterative and explorative trial-an-error integration to enable real time insights into the data. These existing Data Mashup tools are efficient in executing small size data sets, however, they do not emphasize on the run-time efficiency of the data operations. This work is motivated by the limitations of current Data Mashup approaches with regard to data-intensive operations. The run-time of a data operation majorly varies depending on the size of the input data. Hence, in scenarios where one data operation expects inputs from multiple Data Mashup pipelines, which are executed in parallel, a data intensive operation in one of the Data Mashup pipelines leads to a bottleneck, thereby delaying the entire process. The efficiency of such scenarios can be greatly improved by executing the data-intensive operations in a distributed manner. This master thesis copes with this issue through an efficiency optimization of pipeline operators based on Map-Reduce. The Map-Reduce approach enables distributed processing of data to improve the run-time. Map-Reduce is divided into two main steps: (i) the Map step divides a data set into multiple smaller data sets, on which the data operations can be applied in parallel, and (ii) the Reduce step aggregates the results into one data set. The goal of this thesis is to enable a dynamic decision making while selecting suitable implementations for the data operations. This mechanism should be able to dynamically decide, which pipeline operators should be processed in a distributed manner, such as using a Map-Reduce implementation, and which operators should be processed by existing technologies, such as in-memory processing by Web Services. This decision is important because Map-Reduce itself can lead to a significant overhead while processing small data sets. Once it is decided that an operation should be processed using Map-Reduce, corresponding Map-Reduce jobs are invoked that process the data. This dynamic decision making can be achieved through WS-Policies. Web Services use policies to declare in a consistent and standardized manner what they are capable of supporting and which constraints and requirements they impose on their potential requestors. By comparing the capabilities of the Web Service with the requirements of the service requestor, it can be decided if the implementation is suitable for executing the data operation.