Thesis (M.S. in electrical engineering)--Washington State University, May 2010. / Title from PDF title page (viewed on July 15, 2010). "School of Electrical Engineering and Computer Science." Includes bibliographical references (p. 71-75).

Thesis (M.S.S.I.S.)--Regis University, Denver, Colo., 2009. / Title from PDF title page (viewed on Jun. 26, 2010). Includes bibliographical references.

Ο σκοπός αυτής της διατριβής είναι να αναπτυχθούν σχέδια συντονισμού σχετικά με την κίνηση των ρομποτικών πρακτόρων με σκοπό την κάλυψη μιας περιοχής κάτω από RF επικοινωνιακούς περιορισμούς . Οι κόμβοι εκτελούν την κίνηση σε ξεχωριστά χρονικά βήματα σύμφωνα με τις διανεμημένες πληροφορίες που αποκτώνται από τους κόμβους που συνδέονται στον προκαθορισμένο αριθμό hops έως ότου φθάσουν στη βέλτιστη τοπολογία όσον αφορά την κάλυψη της περιοχής. Τα ρομπότ υποτίθεται ότι είναι εξοπλισμένα με έναν αισθητήρα για λόγους κάλυψης και με έναν ράδιο πομποδέκτη έτσι ώστε να μεταδοθούν οι πληροφορίες. Ωστόσο, η ακτίνα επικοινωνίας δεν απαιτείται να είναι τουλάχιστον διπλάσια της ακτίνας του αισθητήρα επισκόπησης, κάτι που προσθέτει έναν πρόσθετο περιορισμό στο γενικό πρόβλημα. Τα σχέδια συντονισμού αναπτύσσονται εξασφαλίζοντας την συνολική RF συνδεσιμότητα του δικτύου επιτυγχάνοντας τη βέλτιστη κάλυψη περιοχής. Τα αποτελέσματα ελέγχονται περαιτέρω μέσω των μελετών προσομοιώσεων. / The purpose of this thesis is to develop coordination schemes concerning the motion of robotic agents for area coverage purposes under RF communications constraints. The nodes perform motion in discrete time steps according to distributed information acquired from nodes which are connected at predefined number of hops until they reach optimum area configuration. Robots are supposed to be equipped with sensor for coverage purposes and with radio transceiver so as information to be transmitted. However, communication radius is not demanded to be at least equal to twice the sensing one, imposing an extra constraint in the overall problem. Coordination schemes are developed ensuring end-to-end RF connectivity of the network while attaining optimum area coverage. Results are further verified via simulations studies.

This thesis addresses the need for a scalable distributed solution for k-nearest-neighbor (kNN) search, a fundamental data mining task. This unsupervised method poses particular challenges on shared-nothing distributed architectures, where global information about the dataset is not available to individual machines. The distance to search for neighbors is not known a priori, and therefore a dynamic data partitioning strategy is required to guarantee that exact kNN can be found autonomously on each machine. Pivot-based partitioning has been shown to facilitate bounding of partitions, however state-of-the-art methods suffer from prohibitive data duplication (upwards of 20x the size of the dataset). In this work an innovative method for solving exact distributed kNN search called PkNN is presented. The key idea is to perform computation over several rounds, leveraging pivot-based data partitioning at each stage. Aggressive data-driven bounds limit communication costs, and a number of optimizations are designed for efficient computation. Experimental study on large real-world data (over 1 billion points) compares PkNN to the state-of-the-art distributed solution, demonstrating that the benefits of additional stages of computation in the PkNN method heavily outweigh the added I/O overhead. PkNN achieves a data duplication rate close to 1, significant speedup over previous solutions, and scales effectively in data cardinality and dimension. PkNN can facilitate distributed solutions to other unsupervised learning methods which rely on kNN search as a critical building block. As one example, a distributed framework for the Local Outlier Factor (LOF) algorithm is given. Testing on large real-world and synthetic data with varying characteristics measures the scalability of PkNN and the distributed LOF framework in data size and dimensionality.

This thesis addresses the need for a scalable distributed solution for k-nearest-neighbor (kNN) search, a fundamental data mining task. This unsupervised method poses particular challenges on shared-nothing distributed architectures, where global information about the dataset is not available to individual machines. The distance to search for neighbors is not known a priori, and therefore a dynamic data partitioning strategy is required to guarantee that exact kNN can be found autonomously on each machine. Pivot-based partitioning has been shown to facilitate bounding of partitions, however state-of-the-art methods suffer from prohibitive data duplication (upwards of 20x the size of the dataset). In this work an innovative method for solving exact distributed kNN search called PkNN is presented. The key idea is to perform computation over several rounds, leveraging pivot-based data partitioning at each stage. Aggressive data-driven bounds limit communication costs, and a number of optimizations are designed for efficient computation. Experimental study on large real-world data (over 1 billion points) compares PkNN to the state-of-the-art distributed solution, demonstrating that the benefits of additional stages of computation in the PkNN method heavily outweigh the added I/O overhead. PkNN achieves a data duplication rate close to 1, significant speedup over previous solutions, and scales effectively in data cardinality and dimension. PkNN can facilitate distributed solutions to other unsupervised learning methods which rely on kNN search as a critical building block. As one example, a distributed framework for the Local Outlier Factor (LOF) algorithm is given. Testing on large real-world and synthetic data with varying characteristics measures the scalability of PkNN and the distributed LOF framework in data size and dimensionality.

The research presented in this dissertation primarily focuses on providing analytical frequency-domain equations that use the system and controller parameters to accurately characterize the power conversion harmonics resulting from the deployment of hysteresis current-controlled inverters within the ac network. In addition, the ac and dc harmonic interactions under both ideal system conditions (constant dc excitation and pure sinusoidal ac voltages) and non-ideal system conditions (harmonic terms are added to the dc and ac sources) are derived for the fixed- and variable-band hysteresis current control. The spectral characteristics, such as frequency orders, spectral magnitude and bandwidth, are given in terms of line and control parameters, a development not only useful in analyzing the harmonic output sensitivity to line and controller parameter variations, but also in filter and system design. Various simulation studies compared results obtained from the developed models to those obtained from the Fourier analysis of MATLAB/Simulink output with very good agreement. The developed models proved their reliability and improved numerical efficiency in harmonic studies compared to those performed using time-domain simulations.

Multi-agent systems rely heavily on coordination and cooperation to achieve a variety of tasks. It is often assumed that these agents will be fully cooperative, or have reliable and equal performance among group members. Instead, we consider cooperation as a spectrum of possible interactions, ranging from performance variations within the group to adversarial agents. This thesis examines several scenarios where cooperation and performance are not guaranteed. Potential applications include sensor coverage, emergency response, wildlife management, tracking, and surveillance. We use geometric methods, such as Voronoi tessellations, for design insight and Lyapunov-based stability theory to analyze our proposed controllers. Performance is verified through simulations and experiments on a variety of ground and aerial robotic platforms. First, we consider the problem of Voronoi-based coverage control, where a group of robots must spread out over an environment to provide coverage. Our approach adapts online to sensing and actuation performance variations with the group. The robots have no prior knowledge of their relative performance, and in a distributed fashion, compensate by assigning weaker robots a smaller portion of the environment. Next, we consider the problem of multi-agent herding, akin to shepherding. Here, a group of dog-like robots must drive a herd of non-cooperative sheep-like agents around the environment. Our key insight in designing the control laws for the herders is to enforce geometrical relationships that allow for the combined system dynamics to reduce to a single nonholonomic vehicle. We also investigate the cooperative pursuit of an evader by a group of quadrotors in an environment with no-fly zones. While the pursuers cannot enter the no-fly zones, the evader moves freely through the zones to avoid capture. Using tools for Voronoi-based coverage control, we provide an algorithm to distribute the pursuers around the zone's boundary and minimize capture time once the evader emerges. Finally, we present an algorithm for the guaranteed capture of multiple evaders by one or more pursuers in a bounded, convex environment. The pursuers utilize properties of the evader's Voronoi cell to choose a control strategy that minimizes the safe-reachable area of the evader, which in turn leads to the evader's capture.

The generation, transmission and distribution of electricity underpins modern technology and constitutes a necessary element for our development and economic functionality. In the recent years, as a result of environmental concerns and technological advances, private and public investment have been steadily turning towards renewable sources of energy, resulting in a growing penetration of those in the power network. This poses additional challenges in the control of power networks, since renewable generation is in general intermittent, and a large penetration may cause frequent deviations between generation and demand, which can harm power quality and even cause blackouts. Load side participation in the power grid is considered by many a means to counterbalance intermittent generation, due to its ability to provide fast response at urgencies. Industrial loads as well as household appliances, may respond to frequency deviations by adjusting their demand in order to support the network. This is backed by the development of relevant sensing and computation technologies. The increasing numbers of local renewable sources of generation along the introduction of controllable loads dramatically increases the number of active elements in the power network, making traditionally implemented, centralised control dicult and costly. This demonstrates the need for the employment of highly distributed schemes in the control of generation and demand. Such schemes need to ensure the smooth and stable operation of the network. Furthermore, an issue of fairness among controllable loads needs to be considered, such that it is ensured that all loads share the burden to support the network evenly and with minimum disruption. We study the dynamic behaviour of power networks within the primary and secondary frequency control timeframes. Using tools from non-linear control and optimisation, we present methods to design distributed control schemes for generation and demand that guarantee stability and fairness in power allocation. Our analysis provides relaxed stability conditions in comparison with current literature and allows the inclusion of practically relevant classes of generation and demand dynamics that have not been considered within this setting, such as of higher order dynamics. Furthermore, fairness in the power allocation between loads is guaranteed by ensuring that the equilibria of the system are solutions to appropriately constructed optimisation problems. It is evident that a synchronising variable is required for optimality to be achieved and frequency is used as such in primary control schemes whereas for secondary frequency control a dierent synchronising variable is adopted. For the latter case, the requirements of the synchronising feedback scheme have been relaxed with the use of an appropriate observer, showing that stability and optimality guarantees are retained. The problem of secondary frequency regulation where ancillary services are provided from switching loads is also considered. Such loads switch on and off when some prescribed frequency threshold is reached in order to support the power network at urgencies. We show that the presence of switching loads does not compromise the stability of the power network and reduces the frequency overshoot, potentially saving the network from collapsing. Furthermore, we explain that when the on and o switching frequencies are equivalent, then arbitrarily fast switching phenomena might occur, something undesirable in practical implementations. As a solution to this problem, hysteresis schemes where the switch on and off frequencies differ are proposed and stability guarantees are provided within this setting.

The need for graph computations is evident in a multitude of use cases. To support computations on large-scale graphs, several parallel systems have been developed. However, existing graph systems require users to recast algorithms into new models, which makes parallel graph computations as a privilege to experienced users only. Moreover, real world applications often require much more complex graph processing workflows than previously evaluated. In response to these challenges, the thesis presents GRAPE, a distributed graph computation system, shipped with various applications for social network analysis, social media marketing and functional dependencies on graphs. Firstly, the thesis presents the foundation of GRAPE. The principled approach of GRAPE is based on partial evaluation and incremental computation. Sequential graph algorithms can be plugged into GRAPE with minor changes, and get parallelized as a whole. The termination and correctness are guaranteed under a monotonic condition. Secondly, as an application on GRAPE, the thesis proposes graph-pattern association rules (GPARs) for social media marketing. GPARs help users discover regularities between entities in social graphs and identify potential customers by exploring social influence. The thesis studies the problem of discovering top-k diversified GPARs and the problem of identifying potential customers with GPARs. Although both are NP- hard, parallel scalable algorithms on GRAPE are developed, which guarantee a polynomial speedup over sequential algorithms with the increase of processors. Thirdly, the thesis proposes quantified graph patterns (QGPs), an extension of graph patterns by supporting simple counting quantifiers on edges. QGPs naturally express universal and existential quantification, numeric and ratio aggregates, as well as negation. The thesis proves that the matching problem of QGPs remains NP-complete in the absence of negation, and is DP-complete for general QGPs. In addition, the thesis introduces quantified graph association rules defined with QGPs, to identify potential customers in social media marketing. Finally, to address the issue of data consistency, the thesis proposes a class of functional dependencies for graphs, referred to as GFDs. GFDs capture both attribute-value dependencies and topological structures of entities. The satisfiability and implication problems for GFDs are studied and proved to be coNP-complete and NP-complete, respectively. The thesis also proves that the validation problem for GFDs is coNP- complete. The parallel algorithms developed on GRAPE verify that GFDs provide an effective approach to detecting inconsistencies in knowledge and social graphs.

This thesis contributes to improve the photovoltaic Distributed Generation (DG) systems by proposing three novel methods to the system. On DC conversion side, a new integrated magnetic structure for interleaved converter and a new method to calculate the eddy current and hysteresis losses in the magnetic core were proposed. On inversion side, A new synchronisation method for grid tie inverters was suggested. The technique is using the Recursive Discrete Fourier Transform (RDFT) to find fundamental in grid waveform. On the DC converter side, the benefits of the new structure is to produce magnetic flux that alternate in the core across both directions of the BH curve. The advantages of alternating magnetic flux are, to increase the Root Mean Square (RMS) value of produced current with respect to core volume that lead to reduce the core size and reducing losses by using high permeability material. Furthermore, the proposed structure led to reduce the number of magnetic components which helped to improve the efficiency. The converter was tested and evaluated were the results show that the topology is able to produce high gain and it shows that the new interleaved structure is efficient. A new method to calculate the eddy current loss was proposed, where the flux waveform in the core was analysed to its original frequency component. Each of the components were utilized individually to find the loss. The effect of changing the duty cycle of the converter was taken into consideration on the total eddy current loss, as it will effect on the total harmonics content in the flux waveform. On the inverter side, due to recent developments combined with the increasing power demand by single phase non-linear loads where voltage spikes, harmonics and DC component were impacted the electric grid quality. These effects can likewise make the synchronisation process a challenge, where filters or Digital Signal processing (DSP) analysers are required to acquire the fundamental component as a consequence to the waveform deformation. A new linear approximation with RDFT is presented in this thesis for grid tie inverters. The new method provides a computation reduction as well as high accuracy in tracking the fundamental frequency in a distorted grid during synchronisation. The method accuracy was proved mathematically and simulated with different input signals. Error in magnitude and frequency measurement were measured, presented and compared with other research in order to verify the proposed method.