abstract: Visual applications – those that use camera frames as part of the application – provide a rich, context-aware experience. The continued development of mixed and augmented reality (MR/AR) computing environments furthers the richness of this experience by providing applications a continuous vision experience, where visual information continuously provides context for applications and the real world is augmented by the virtual. To understand user privacy concerns in continuous vision computing environments, this work studies three MR/AR applications (augmented markers, augmented faces, and text capture) to show that in a modern mobile system, the typical user is exposed to potential mass collection of sensitive information, posing privacy and security deficiencies to be addressed in future systems. To address such deficiencies, a development framework is proposed that provides resource isolation between user information contained in camera frames and application access to the network. The design is implemented using existing system utilities as a proof of concept on the Android operating system and demonstrates its viability with a modern state-of-the-art augmented reality library and several augmented reality applications. Evaluation is conducted on the design on a Samsung Galaxy S8 phone by comparing the applications from the case study with modified versions which better protect user privacy. Early results show that the new design efficiently protects users against data collection in MR/AR applications with less than 0.7% performance overhead. / Dissertation/Thesis / Masters Thesis Computer Engineering 2019

none

Today's telecommunications networks are relying more and more on optical fibers as their physical medium. Currently the Wavelength Division Multiplexing technology enables hundreds of wavelengths to be multiplexed on a single fiber. Using this technology capacity can be dramatically increased, even to the order of Terabits per second. While WDM technology has given a satisfactory answer to the ever-increasing demand for capacity, there is still a problem which needs to be handled efficiently: survivability. Our proposed fault restoration system optimized between restoration cost and speed. We extended the concept of Forward Equivalence Class (FEC) in Multi Protocol Label switching (MPLS) to our proposed fault restoration system. Speed was found to be in the order of 1 to 3 microseconds using predesigned protection, depending on the configuration of the system. Optimization was done between restoration speed and cost by introducing a priority field in the packet header.

The main purpose of this research is the implementation of a software interface. This interface shall allow the interpretation of particle size in a medium with respect to its diffraction patterns. The literature shows extensive work on the theory of light scattering but the experiments are cumbersome to implement. Some initial work has required the levitation of particle to isolate the difficulties associated with a flow environment. The purpose of this work; however, will focus on the software requirements to synchronize, collect and analyze light scattering patterns. Although there are many other ways of sizing particles, it may be useful to prove the feasibility of the well-defined theory in a flow environment. The light scattering signatures from an illuminated particle is abundantly used in the flow cytometry area but are obtained from other mean in capturing light information. The present study could determine the specific angles of interest allowing the discrimination by size of various types of particles (e.g.. blood cells).

The ever-increasing growth of the demand in IT computing, storage and large-scale cloud services leads to the proliferation of data centers that consist of (tens of) thousands of servers. As a result, data centers are now among the largest electricity consumers worldwide. Data center energy and resource efficiency has started to receive significant attention due to its economical, environmental, and performance impacts. In tandem, facing increasing challenges in stabilizing the power grids due to growing needs of intermittent renewable energy integration, power market operators have started to offer a number of demand response (DR) opportunities for energy consumers (such as data centers) to receive credits by modulating their power consumption dynamically following specific requirements. This dissertation claims that data centers have strong capabilities to emerge as major enablers of substantial electricity integration from renewables. The participation of data centers into emerging DR, such as regulation service reserves (RSRs), enables the growth of the data center in a sustainable, environmentally neutral, or even beneficial way, while also significantly reducing data center electricity costs. In this dissertation, we first model data center participation in DR, and then propose runtime policies to dynamically modulate data center power in response to independent system operator (ISO) requests, leveraging advanced server power and workload management techniques. We also propose energy and reserve bidding strategies to minimize the data center energy cost. Our results demonstrate that a typical data center can achieve up to 44% monetary savings in its electricity cost with RSR provision, dramatically surpassing savings achieved by traditional energy management strategies. In addition, we investigate the capabilities and benefits of various types of energy storage devices (ESDs) in DR. Finally, we demonstrate RSR provision in practice on a real server. In addition to its contributions on improving data center energy efficiency, this dissertation also proposes a novel method to address data center management efficiency. We propose an intelligent system analytics approach, "discovery by example", which leverages fingerprinting and machine learning methods to automatically discover software and system changes. Our approach eases runtime data center introspection and reduces the cost of system management. / 2018-11-04T00:00:00Z

One of the unique characteristics of wireless sensor networks (WSNs) is that sensor nodes have very constrained resources. Typical sensor nodes have lower computing power, communication bandwidth, and smaller memory than other wireless devices, and operate on limited capacity batteries. Hence power efficiency is very important in WSNs because power failure of some sensor nodes may lead to total network failure. In many cases the WSNs have to operate in harsh environments without human intervention for expended period time. Thus, much research on reducing or minimizing the power consumption, and thereby increasing the network lifetime, has been performed at each layer of the network layers. In this dissertation we approach three important issues related power management in WSNs: routing, time synchronization, and medium access control (MAC). We first discuss the effect of selecting routing protocols on the lifetime of the WSNs. The maximum and minimum bounds of the lifetime with respect to the routing protocols are derived. The routing protocols corresponding to the bounds are also presented. The simulation results show that the choice of the routing protocol has very little impact on the lifetime of the network and that simple routing protocols such as shortest path routing perform very close to the the maximum bound of the lifetime of the network. Next, we propose a simple and accurate time synchronization protocol that can be used a a fundamental component of other synchronization-based protocols in WSNs. Analytical bounds on the synchronization errors of proposed protocol are discussed. The implementation results on Mica2 and Telos motes show that proposed time synchronization protocol outperforms existing ones in terms of the precision and required resources. Finally, we model the power consumption of WSN MAC protocols. We derive analytically the power consumption of well known MAC protocols for WSNs, and analyze and compare their performance. We validate the models by measuring the power consumption on Mica 2 motes and comparing those measured power consumption with the analytical results.

BURKE, DAVID ALEXANDER. A conservative approach to mounting and applying an omni-directional vision system onto EvBot II mobile robot platforms, for use in accurate formation control (Under the direction of Edward Grant). The research sensing capabilities of the EvBot II mobile robot platforms were increased and enhanced by the addition of the omni-directional camera. This, along with the associated machine vision capabilities maintained the conservative approach of the EvBot II philosophy, fiscal responsibility with computational optimality. The research increased the capabilities of the EvBot II platform by demonstrating that omni-directional vision processing could be performed relatively economically on a PC 104, while leaving as much processor time available as possible for running other programs.

The purpose of this research has been to develop methods of improving the generalization capabilities of artificial neural networks. Tools for examining the influence of individual training set patterns on the learning abilities of individual neurons are put forth and utilized in the implementation of new network learning algorithms. Algorithms are based largely on the supervised training algorithm: backpropagation, and all experiments use the standard backpropagation algorithm for comparison of results. The focus of the new learning algorithms revolve around the addition of two main components. The first addition is that of an unsupervised learning algorithm called flocking. Flocking attempts to provide network hyperplane divisions that are evenly influenced by examples on either side of the hyperplane. The second addition is that of a multi-tasking approach called convergence training. Convergence training uses the information provided by a clustering algorithm in order to create subtasks that represent the divisions between clusters. These subtasks are then trained in unison in order to promote hyperplane sharing within the problem space. Generalization was improved in most cases and the solutions produced by the new learning algorithms are demonstrated to be very robust against different random weight initializations. This research is not only a search for better generalizing ANN learning algorithms, but also a search for better understanding when dealing with the complexities involved in ANN generalization.

The research presented in this thesis describes the development of the Linux distribution and a new control architecture for robots. The reasons Linux was chosen are enumerated and a description of the build system and setup used to generate the distribution, with support for multiple platforms, is discussed. The Evbot Abstraction Layer (EAL), a new robot control architecture and framework is described, and the simple API is detailed.

Value prediction is a speculative technique to break true data dependencies by predicting uncomputed values based on history. Previous research focused on exploiting two types of value locality (computation-based and context-based) in the local value history, which is the value sequence produced by the same instruction that is being predicted. Besides local value history, value locality also exists in global value history, which is the value sequence produced by all dynamic instructions according to their execution order. In this thesis, a new type of value locality, computational locality in global value history is studied. A prediction scheme, called gDiff, is designed to exploit one special and most common case of this computational model, the stride-based computation, in global value history. Experiments show that there exists very strong stride type of locality in global value sequences and ideally the gDiff predictor can achieve 73% prediction accuracy for all value producing instructions without any hybrid scheme, much higher than local stride and local context prediction schemes. However, the ability to realistically exploit locality in global value history is greatly challenged by the value delay issue, i.e., the correlated value may not be available when the prediction is being made. The value delay issue is studied in an out-of-order (OOO) execution pipeline model and the gDiff predictor is improved by maintaining an order in the value queue and utilizing local stride predictions when global values are unavailable to avoid the value delay problem. This improved predictor, called hgDiff, demonstrates 88% accuracy and 69% prediction coverage on average, outperforming a local stride predictor by 2% higher accuracy and 13% higher coverage.