Process optimization is used to improve the utility and the economic performance of industrial processes, and is as such central in most automation strategies. In this thesis, two feedback-based methods for online process optimization are considered: Extremum seeking control (ESC), a classic model-free method used for steady-state optimization which dates back to the early 1900's, and economic model predictive control (EMPC), a more recent method which utilizes a model to dynamically optimize the closed-loop process economics in real time. Part I of the thesis concerns ESC. Due to a well known result by Krsti\'c and Wang, it is known that the classic ESC-loop will possess a stable stationary solution in a neighborhood of the optimum when applied to dynamic plants. However, existence and stability of an optimal solution alone are not sufficient to guarantee that the ESC-loop will converge to the optimum; uniqueness also has to be considered. In this thesis, it is shown that the near-optimal solution is not necessarily unique, not even in cases where the objective, i.e., the steady-state input-output map, is convex. The stationary solutions to the loop are shown to be characterized by a condition on the local plant phase-lag, and for a biochemical reactor it is found that this condition can be satisfied not only locally at the optimum but also at arbitrary points away from the optimum. Bifurcation theory is used to show that the observed solution multiplicity may be explained by existence of fold bifurcation points, and conditions for existence of such points are given. The phase-lag condition for stationarity combined with the result by Krsti\'c and Wang suggest that the process phase-lag is connected to steady-state optimality. In this thesis, it is shown that the steady-state optimum corresponds to a bifurcation of the plant zero dynamics which is reflected in large local phase-lag variations. This explains why the classical ESC method will have a near-optimal stationary solution when applied to dynamic plants, and it also shows that a steady-state optimum may be located using only phase information. Finally, we introduce greedy ESC which is applicable to plants where the dynamics may be separated into different time-scales. By optimizing only the fast plant-dynamics, significant performance improvements may be achieved. Part II of this thesis concerns EMPC. The method is first evaluated for optimization of a paper-making process by means of simulations. These reveal several important properties of EMPC, e.g., that EMPC in the presence of excessive degrees of freedom automatically selects the inputs which are currently most efficient, and that EMPC effectively plans ahead which leads to significantly improved performance during grade changes. However, it is also observed that EMPC often operates with constraints active since economic objectives frequently are monotone, and this may lead to issues with robustness. To avoid active constraints, constraint margins are introduced to force the closed-loop to operate in the interior of the feasible set. The margins affect the economic performance significantly and the optimal choice is dependent on the uncertainty present in the problem. To avoid modeling of the uncertainty, it is suggested that the margins are adapted based on feedback from the realized closed-loop economic performance. / <p>QC 20180829</p>

The present thesis is concerned with the problem of robust pose estimation for planar targets in the context of real-time mobile vision. As a consequence of this research, individual developments made in isolation by earlier researchers are here considered together. Several adaptations to the existing algorithms are undertaken yielding a unified framework for robust pose estimation. This framework is specifically designed to meet the growing demand for fast and robust estimation on power-constrained platforms. For robust recognition of targets at very low computational costs, we employ feature based methods which are based on local binary descriptors allowing fast feature matching at run-time. The matching set is then fed to a robust parameter estimation algorithm in order to obtain a reliable homography. On the basis of our experimental results, it can be concluded that reliable homography estimates can be obtained using a device-friendly implementation of the Gaussian Elimination algorithm. We also show in this thesis that our simplified approach can significantly improve the homography estimation step in a hypothesize-and-verify scheme. The author's attention is focused not only on developing fast algorithms for the recognition framework but also on the optimized implementation of such algorithms. Any other recognition framework would similarly benefit from our optimized implementation.

Due to the need for advanced computer-controlled functionality in distributed embedded systems the requirements on network communication are becoming overly intricate. This dissertation targets the requirements that are concerned with real-time guarantees, run-time adaptation, resource utilization and flexibility during the development. The Flexible Time-Triggered Switched Ethernet (FTT-SE) and Hard Real-Time Ethernet Switching (HaRTES) network architectures have emerged as two promising solutions that can cater for these requirements. However, these architectures do not support multi-hop communication as they are originally developed for single-switch networks. This dissertation presents a fundamental contribution in multi-hop real-time communication over the FTT-SE and HaRTES architectures targeting the above mentioned requirements. It proposes and evaluates various solutions for scheduling and forwarding the traffic through multiple switches in these architectures. These solutions preserve the ability of dynamic adaptation without jeopardizing real-time properties of the architectures. Moreover, the dissertation presents schedulability analyses for the timeliness verification and evaluation of the proposed solutions as well as several protocols to support run-time adaptation in the multi-hop communication. Finally, the work led to an end-to-end resource reservation framework, based on the proposed multi-hop architectures, to support flexibility during the development of the systems. The efficiency of the proposed solutions is evaluated on various case studies that are inspired from industrial systems.

This work deals with the problem of accurately rendering mirror reflections on curved surfaces in real-time. While planar mirrors do not pose a problem in this area, non-planar surfaces are nowadays rendered mostly using environment mapping, which is a method of approximating the reflections well enough for the human eye. However, this approach may not be suitable for applications such as CAD systems. Accurate mirror reflections can be rendered with ray tracing methods, but not in real-time and therefore without offering interactivity. This work examines existing approaches to the problem and proposes a new algorithm for computing accurate mirror reflections in real-time using accelerated searching for intersections with depth profile stored in cubemap textures. This algorithm has been implemented using OpenGL and tested on different platforms.

Diplomová práce se zabývá návrhem a realizací vestavěného řídicího systému určeného pro autonomní mobilní robot Advee. Řídicí systém tvoří vrstvu abstrakce mezi hardwarovými prostředky robotu a vyššími vrstvami řízení, které provádějí lokalizaci robotu a plánování pohybu. V rámci návrhu byla vyvinuta modulární struktura systému a zvoleny prostředky mezimodulové komunikace. Navržený systém byl pak implementován včetně podpory komunikačních standardů EIA-485 a CAN bus. Zvolená architektura systému se v praxi osvědčila --- prototyp robotu Advee řízený popsaným systémem má za sebou více než 500 hodin komerčního provozu s minimem poruch.

<p>The major objective of this project is to estimate the resource requirements of a real-time program developed in the actor-based system ACT++. This report describes the design and implementation of instrumentation primitives. Using these primitives, a real-time system can be simulated and an estimate of processor and memory requirements can be obtained.</p> / Master of Science

Jedes Jahr werden am Universitätsklinikum Würzburg etwa 100 PatientInnen allogen stammzelltransplantiert. Für den Erfolg einer allogenen Stammzelltransplantation ist neben einer passenden Spenderauswahl und einer optimalen Vorbereitung auch die regelmäßige Nachkontrolle wichtig. Im Rahmen dieser Nachkontrollen werden unter anderem Chimärismusuntersuchungen durchgeführt. Als Chimäre wird in der Wissenschaft ein Lebewesen bezeichnet, das Zellen in sich trägt, die aus zwei oder mehr Zygoten entstanden ist. Der Begriff kommt aus der griechischen Mythologie, in der die Chimäre als Mischwesen aus Löwe, Ziege und Schlange (oder manchmal Drache) beschrieben wurde. Lange Zeit galt die STR/VNTR Methode als Goldstandard für die Nachkontrolle der Chimärismusuntersuchung. Durch Alizadeh et al. wurde bereits im Jahr 2002 eine neuartige Methode beschrieben, die mittels RT-PCR den genauen Chimärismusgrad messen kann. Diese Arbeit beschäftigte sich mit der Etablierung und Evaluierung dieser Methode der Chimärismusuntersuchung am Universitätsklinikum Würzburg. Es konnte gezeigt werden, dass sie für den klinischen Alltag geeignet ist und wurde auch im Hinblick auf äußere Störfaktoren untersucht. Darüber hinaus wurde eine Methode zur Bestimmung des CD3+ Chimärismus etabliert. Diese Untersuchung ist wichtig für die Abschätzung eines möglichen Abstoßungs- und GvHD Risikos. Seit 2013 finden auf Basis dieser Arbeit regelmäßige Chimärismusuntersuchungen am Universitätsklinikum Würzburg statt. / Every year, hematopoietic stem cell transplantations are being performed on approximately 100 patients at the university hospital Würzburg. Apart from finding a fitting stem cell donor and a perfect preparation of the recipient, follow-up examinations are essential for the long-term success of this procedure. Monitoring the chimerism after allogeneic stem cell transplantation is an essentialpart of those follow-up examinations. In science, a chimera is defined as a living organism, that contains cells deriving from different zygotes. In greek mythology, the „Chimera“ was described as a creature with the head of a lion, the body of a goat and the tail of a snake. For many years, the STR/VNTR method was the standard for post-transplant chimerism testing. In 2002, Alizadeh et al. developed a new method using RT-PCR to monitor chimerism. This dissertation evaluated and established this new method for engraftment analysis of transplanted patients at the university hospital Würzburg. It could be shown that the method was suitable for routine use in clinical practice. In addition, a method to measure CD3 + chimerism was established to evaluate the risks of GvHD or graft failure. Since 2013, chimerism is being tested at the university hospital Würzburg based on the results of this graduate thesis.

This research focuses on the development of a novel adaptive dynamical system approach to vector quantization or clustering based on only ordinary differential equations (ODEs) with potential for a real-time implementation. The ODE-based approach has an advantage in making it possible real-time implementation of the system with either electronic or photonic analog devices. This dynamical system consists of a set of energy functions which create valleys for representing clusters. Each valley represents a cluster of similar input patterns. The proposed system includes a dynamic parameter, called vigilance parameter. This parameter approximately reflects the radius of the generated valleys. Through several examples of different pattern clusters, it is shown that the model can successfully quantize/cluster these types of input patterns. Also, a hardware implementation by photonic and/or electronic analog devices is given In addition, we analyze and study stability of our dynamical system. By discovering the equilibrium points for certain input patterns and analyzing their stability, we have shown the quantizing behavior of the system with respect to its parameters. We also extend our model to include competition mechanism and vigilance dynamics. The competition mechanism causes only one label to be assigned to a group of patterns. The vigilance dynamics adjust vigilance parameter so that the cluster size or the quantizing resolution can be adaptive to the density and distribution of the input patterns. This reduces the burden of re-tuning the vigilance parameter for a given input pattern set and also better represents the input pattern space. The vigilance parameter approximately reflects the radius of the generated valley for each cluster. Making this parameter dynamic allows the bigger cluster to have a bigger radius and as a result a better cluster. Furthermore, an alternative dynamical system to our proposed system is also introduced. This system utilizes sigmoid and competitive functions. Although the results of this system are encouraging, the use of sigmoid function makes analyze and study stability of the system extremely difficult.

Flavobacterium columnare, a Gram-negative rod-shaped bacterium, is the causative agent of columnaris disease in a variety of fish hosts but is of particular significance to the catfish industry located in the southeastern United States. Columnaris infections are a leading cause of mortalities in catfish ponds, occurring alone or in conjunction with other diseases. Typical diagnostic methods for columnaris infections involve the use of selective media following the observation of gross signs of disease. A real-time quantitative PCR (qPCR) assay to estimate the quantity of bacteria present in environmental and tissue samples was developed and validated. The genetic variability seen in F. columnare makes detection of isolates from different genomovars (genetic groups) essential to an assay for diagnostic application. Isolates from catfish generally fall into one of two different genomovars, one being virulent to catfish, while the other genomovar is thought to be largely opportunistic. The qPCR assay described herein was designed specifically to detect F. columnare isolates from the two major genomovars most often associated with farm-raised catfish. The assay was shown specific to F. columnare, regardless of genomovar, and demonstrated sensitivity consistent with similar qPCR assays. In addition, the assay provides quantitative information, estimating the bacterial loads in fish tissue and the environment. Two different applications of the assay are presented: (1) Estimate bacterial burden in fish tissue following immersion challenges to identify variation in transmission rates between channel and blue x channel hybrid catfish, and (2) Estimate the environmental burden of F. columnare in catfish ponds over the course of a single calendar year. This assay will provide an invaluable tool for researchers and diagnosticians in expanding our understanding of F. columnare and how it interacts with the host and environment.

Disruptions of the memory systems in the brain are linked to the manifestation of many neuropsychiatric diseases such as Alzheimer’s disease, depression, and post-traumatic stress disorder. The limited efficacy of current treatments necessities the development of more effective therapies. Neuromodulation has proven effective in a variety of neurological diseases and could be an attractive solution for memory disorders. However, the application of neuromodulation requires a more detailed understanding of the network dynamics associated with memory formation and recall. In this work, we applied a combination of optical and computational tools in the development of a novel strategy for the modulation of memories, and have expanded its application for interrogation of the hippocampal circuitry underlying memory processing in mice. First, we developed a closed-loop optogenetic stimulation platform to activate neurons implicated in memory processing (engram neurons) with a high temporal resolution. We applied this platform to modulate the activity of engram neurons and assess memory processing with respect to synchronous network activity. The results of our investigation support the proposal that encoding new information and recalling stored memories occur during distinct epochs of hippocampal network-wide oscillations. Having established the high efficacy of the modulation of engram neurons’ activity in a closed-loop fashion, we sought to combine it with two-photon imaging to enable high spatial resolution interrogation of hippocampal circuitry. We developed a behavioral apparatus for head-fixed engram modulation and the assessment of memory recall in immobile animals. Moreover, through the optimization of dual color two-photon imaging, we improved the ability to monitor activity of neurons in the subfields of the hippocampus with cellular specificity. The platform created here will be applied to investigate the effects of engram reactivation on downstream projections targets with high spatial and cell subtype specificity. Following these lines of investigations will enhance our understanding of memory modulation and could lead to novel neuromodulation treatments for neurological disorders associated with memory malfunctioning.