Promotionen der Fakultät für Informatik im Jahr 2022

06.10.2022

Rashid, Aquib
"Towards Safe, Efficient and Flexible Human Robot Collaboration with Traditional Industrial Robots using Local and Global Sensing"
Promotion zum Dr.-Ing.

Abstract:

Human robot collaboration is need of the hour as a cost effective solution for agile assembly and disassembly tasks in various industries. Traditional industrial robots (TIR), with large range and higher payload capabilities better complement human worker. However, their high inherent weight and high speeds make them dangeous in the collision event. This requires contactless collision detection and avoidance strategies. Optical sensor based depth sensing system exist but trigger robot-stop on human entrance. However, for human robot collaborative processes with constant human presence, such sensing strategies result in reduced process efficiency.

This work aims to setup not only safe but efficient collaboraiton with a human operator. The framework provides methodology of setting up of the robot cell using flexible sensor concepts and efficient multi-sensor calibration with robot. This ensures  ease in the setup of multisensor system in industrial robot cells, while saving more than 50% of data collection time compared to individual dual system calibration. The work proposes an integrated vision and control system, which efficiently determines the minimum distance to possible human operator. The method for collision sensitivity as well as controller reactivity has been proposed and validated. Collision sensing is performed using local to the robot body- and global to the robot cell- depth measuremnets. Collision avoidance selects one of the corresponding system configurations based on human localization. This enables safe acceleration of the robot compared to the state of technology, and enables 11% lesser process time compared to state of technology

16.09.2022

Reißner, Michael
"Rabattprobleme aus Konsumentensicht Eine Online- und Offlineanalyse"
Promotion zum Dr.rer.nat.

Abstract (DE):

Einem Konsumenten werden in verschiedensten Situationen Rabatte angeboten. In dieser Dissertation wird die Frage untersucht, wie solche Rabattsituationen aus Konsumentensicht formalisiert und Kaufentscheidungen getroffen werden können. Dahingehend wird eine neue Gruppe von acht Problemen, die auf alltäglichen Erfahrungen mit Rabattaktionen basieren, eingeführt. Diese Probleme sind hinsichtlich der Rabattgrundlage (Stempel / Punkte), des Kartentyps (Einzelkarte, Gruppenkarte) und der Frage, ob Stempel beziehungsweise Punkte für Käufe mit Rabatt gesammelt werden, unterschieden. Der inhärenten Planungsunsicherheit für Konsumentenentscheidungen wird explizit durch die Betrachtung jedes Problems als Onlinesituation begegnet. Jedes der acht Probleme wird als Entscheidungs-, Optimierungs- und Onlineproblem beleuchtet.

Abstract (EN):

A consumer is offered discounts in a variety of situations. The central question investigated in this dissertation is how to formalize such discount situations from a consumer perspective, and what methods for deducing purchase decisions are possible. Therefore, a new group of eight discount problems based on everyday experience with loyalty programs is introduced. These problems are distinguished by discount basis (stamps / points), card type (single / group) and whether stamps or points are collectable if a discount is granted. The inherent uncertainty in consumer decisions is explicitly taken into account by considering each of these problems as an online situation as well. Each of the eight problems is examined as a decision, an optimization and an online problem.

14.07.2022

Mahdiani, Mitra
"Advanced Scheduling Techniques for Mixed-Criticality Systems"
Promotion zum Dr.-Ing.

Abstract:

Typically, a real-time system consists of a controlling system (i.e., a computer) and a controlled system (i.e., the environment). Real-time systems are those systems where correctness depends on two aspects: i) the logical result of computation and, ii) the time in which results are produced. It is essential to guarantee meeting timing constraints for this kind of systems to operate correctly. Missing deadlines in many cases — in so-called hard real-time systems — is associated with economic loss or loss of human lives and must be avoided under all circumstances.
On the other hand, there is a trend towards consolidating software functions onto fewer processors in different domains such as automotive systems and avionics with the aim of reducing costs and complexity. Hence, applications with different levels of criticality that used to run in isolation now start sharing processors. As a result, there is a need for techniques that allow designing such mixed-criticality (MC) systems — i.e., real-time systems combining different levels of criticality — and, at the same time, complying with certification requirements in the different domains.
In this research, we study the problem of scheduling MC tasks under EDF (Earliest Deadline First) and propose new approaches to improve scheduling techniques. In particular, we consider that a mix of low-criticality (LO) and high-criticality (HI) tasks are scheduled on one processor. While LO tasks can be modeled by minimum inter-arrival time, deadline, and worst case execution time (WCET), HI tasks are characterized by two WCET parameters: an optimistic and a conservative one.
Basically, the system operates in two modes: LO and HI mode. In LO mode, HI tasks run for no longer than their optimistic execution budgets and are scheduled together with the LO tasks. The system switches to HI mode when one or more HI tasks run for more than their conservative execution budgets. In this case, LO tasks are immediately discarded so as to be able of accommodating the increase in HI execution demand. We propose an exact test for mixed-criticality EDF, which increases efficiency and reliability when compared with the existing approaches from the literature. On this basis, we further derive approximated tests with less complexity and, hence, a reduced running time that makes them more suitable for online checks.

27.04.2022

Peitek, Norman
"A Neuro-Cognitive Perspective of Program Comprehension"
Promotion zum Dr.rer.nat.

Abstract:

Program comprehension is the cognitive process of understanding source code. Since program comprehension is an internal cognitive process, it is inherently difficult to observe and measure reliably. Novel neuroimaging measures, such as functional magnetic resonance imaging (fMRI), provide an additional perspective on program comprehension which promise new insights to program comprehension. This dissertation develops a framework for studying program comprehension with neuroimaging, psycho-physiological, eye tracking, and behavioral methods. Our framework offers a detailed, multi-modal view on program comprehension that allows us to examine even small effects. We shed light on the underlying cognitive process of program comprehension by applying our experiment framework. One major focus is to understand experienced programmers' efficient top-down comprehension. We also link programmers' cognition to common code complexity metrics. In our interdisciplinary research, we show how neuroimaging methods, such as fMRI, in combination with psycho-physiological, eye tracking and behavioral measures, is beneficial to software-engineering research. This dissertation provides a foundation to further investigate the neuro-cognitive perspective to programmers' brains, which is a critical contribution to the future of software engineering.

26.04.2022

Dietze, Robert
"Suchbasierte Algorithmen für das Scheduling unabhängiger paralleler Tasks"
Promotion zum Dr.-Ing.

Abstract:

In parallelen Anwendungen, die auf Grundlage des Programmiermodells >> der gemischten Parallelität implementiert wurden, lassen sich meist unabhängige >> Programmteile (Tasks) identifizieren, die sowohl parallel zueinander als auch selbst >> parallel ausgeführt werden können. Zur Reduzierung der Ausführungszeit solcher Anwendungen auf einem >> parallelen System wird eine zeitliche und räumliche Zuordnung dieser parallelen Tasks >> zu den Prozessoren benötigt, welche mithilfe von Schedulingverfahren ermittelt werden kann. Jedoch ist bereits das Scheduling voneinander abhängiger >> Single-Prozessor-Tasks auf ein paralleles System mit zwei Prozessoren NP-schwer, weshalb zur Lösung von Schedulingproblemen häufig List-Scheduling-Heuristiken verwendet werden. Das Scheduling unabhängiger paralleler Tasks ist aufgrund der >> vielen zusätzlichen Zuordnungsmöglichkeiten deutlich komplexer und erfordert daher >> dedizierte Lösungsverfahren.

Einen vielversprechenden Ansatz zur Lösung komplexer >> Schedulingprobleme bilden suchbasierte Verfahren, die lokale oder globale Suchstrategien zur Lösungsfindung nutzen. In der vorliegenden Arbeit wird untersucht, inwieweit sich >> derartige Verfahren für das Scheduling unabhängiger paralleler Tasks auf heterogene Systeme >> bestehend aus Multicore-Rechnern mit unterschiedlichen Eigenschaften eignen. Zu diesem Zweck werden vier suchbasierte Schedulingverfahren >> entwickelt und untersucht. Konkret werden zwei modifizierende und zwei inkrementelle Verfahren >> vorgestellt, die von Suchverfahren wie der A*-Suche und Metaheuristiken wie der >> Tabu-Suche und des Simulated Annealing inspiriert sind. Zusätzlich wird eine Kostenmodellierung in Form von >> parametrisierten Laufzeitformeln präsentiert, mit der die Ausführungszeiten der parallelen Tasks auf >> heterogenen Systemenmodelliert werden können. Die Verfahren werden in Laufzeitmessungen auf heterogenen >> Rechnerplattformen untereinander und mit existierenden List-Scheduling-Heuristiken verglichen. Als Anwendungen für die Messungen werden sowohl Programme der >> SPLASH-3-Benchmark-Suite als auch eine praxisnahe Simulationsanwendung zur Bauteilbelastung untersucht. Die Ergebnisse zeigen, dass alle vier Verfahren im Vergleich zu >> existierenden List-Scheduling-Heuristiken eine signifikante Reduktion der Ausführungszeit >> erreichen können.

10.02.2022

Ullrich, Markus
"Performance Modelling for Optimized Resource Management and Application Deployment in Cloud Environments"
Promotion zum Dr.-Ing.

Abstract:

Cloud computing is an exciting concept that propels the development of technologies, the creation and expansion of businesses and the rapid prototyping of new ideas. Utilizing the advantages the cloud offers to their fullest potential is not a simple task and thus often users struggle with the technological aspects, lose revenue or do not attempt to benefit from this idea at all. In this dissertation, we identify the lack of standards for performance descriptions as well as the steep learning curve to get familiar with the cloud, which is further amplified by the abundance of available services, as the most prevalent issues that individuals and companies encounter. We further show the relevance of solving these issues by outlining the expected impact, which includes decreased time and financial detriments for individuals and companies as well as a negative effect on the environment. To solve the identified problems we propose the development of a cloud broker with three key components that utilize a performance oriented resource and application model to 1) compare arbitrary resources and applications in a fair manner based on general information, collected with standard benchmark tools 2) select the optimal infrastructure for any application by estimating its resource consumption and execution time and 3) automatically create and manage the selected infrastructure as well as the application deployment. Our contributions to this proposal include the development and test of prototypical proof-of-concept implementations for the three components, the design of the underlying resource and application performance model as well as the selection of appropriate, generic benchmark solutions, which we deployed on two major public clouds using our prototypes. In an extensive objective-based evaluation we assess that we contributed towards solving all the major issues that we identified to increase the usability and efficiency of cloud computing by enabling a better comprehension of resource and application performance in cloud environments and by reducing the necessary time and effort to deploy arbitrary applications in the cloud. We conclude by interpreting the evaluation results and providing an outlook towards future work.

24.01.2022

Englisch, Norbert
"Entwicklungs- und Testunterstützung für Steuergeräte mit AUTOSAR Architektur"
Promotion zum Dr.-Ing.

Abstract: