ECE PhD Graduates

Photo: Bhattarai Manish

Manish Bhattarai [Fall 2020]
11-05-2020

Biography

Mr. Manish Bhattarai defended his PhD thesis on Thursday, November 5 at 10:30 am in a Zoom meeting chaired by Dr. Manel Martinez-Ramon. The title of Mr. Bhattarai's dissertation was "Integrating deep learning and augmented reality to enhance situational awareness in firefighting environment."

Abstract

Intelligent detection and processing capabilities can improve the safety, efficiency, and successful completion of rescue missions conducted by firefighters in emergency first response settings. This research aims to enhance firefighters’ situational awareness by creating an automated system that combines the latest state-of-the-art deep learning networks and infrared imagery gathered on the scene by firefighters. The newly developed algorithm is capable of real-time, intelligent object detection, recognition, tracking, and segmentation. The processed results are used as inputs in the creation of an augmented reality capable of advising firefighters of their location and key features around them, vital to the rescue operation at hand, as well as a path planning feature that acts as a virtual guide to assist disoriented first responders in getting back to safety.

First, we used a deep Convolutional Neural Network (CNN) system to classify and identify objects of interest from thermal imagery in real-time, which can accurately inform firefighters’ decision-making process with up-to-date scene information by extracting, processing, and analyzing crucial information.

Second, we extended this CNN framework for object detection, tracking, and segmentation with a Mask RCNN framework. Utilizing the new information pro- duced by the framework, firefighters can make more informed inferences about their environment that will improve those decisions that are crucial to their safe naviga- tion through such hazardous and potentially catastrophic environments. We take this system further by combining the object recognition framework with an Natural Language Processing(NLP) model for caption generation to provide the firefighter with descriptions of important scene characteristics gathered by the CNN.

Third, we built a deep Q-learning-based agent, immune to stress-induced disori- entation and anxiety, which was able to make clear navigation decisions based on the observed and stored facts in live-fire environments. As a proof of concept, we imitated structural fire in a gaming engine called Unreal Engine. This virtual setting enabled accelerated learning of the environmental conditions through the agent’s in- teraction with the simulated fire conditions built into the game space. The agent is trained with a deep Q-learning algorithm based on a set of rewards and penalties derived from the gaming agent’s actions on the environment.

Fourth and finally, we use a low computational unsupervised learning technique called tensor decomposition to perform meaningful feature extraction in real-time. This method allows for rapid assessment of anomalous events recorded by the camera that can be processed much more quickly than data interpreted through the CNN. The information gained from this technique can be used as a warning system for firefighters to alert them to rapidly changing conditions such as room temperature or smoke prevalence. The method can be deployed with lower power, memory and computational requirements but achieve similar accuracy. Tensor decomposition aid the recognition process by decomposing the sensor data along each dimension for blind source separation and anomaly detection.

When combined, these four approaches present a new approach to information understanding, transfer, and synthesis that could dramatically improve firefighter response and efficacy and reduce the loss of life.

Photo: Andrea Schmidt

Andrea Schmidt [Fall 2020]
09-13-2020

Biography

Ms. Andrea Schmidt defended his PhD thesis on Friday, November 13 at 9 am in a Zoom meeting chaired by Dr. Edl Schamiloglu. The title of Ms. Schmidt's dissertation was "Theoretical and Experimental Studies of the Emission of Electromagnetic Radiation by Superluminal Polarization Currents."

Abstract

Recent experiments, conducted at Los Alamos National Laboratory, have shown that polarization currents (i.e., ∂P/∂t, where P is formally defined as the dipole moment per unit volume), accelerated to superluminal speeds within a dielectric solid, emit tightly focused packets of electromagnetic radiation. Extended faster-than-light sources are distinct from emitters that employ surface currents of free electrons on localized elements such as dipoles to produce radiation. They are true volume sources whose radiation characteristics are entirely dependent on the speed and acceleration of the moving polarization along with the size, shape, and dielectric constant of the solid that contains it. Prototype general-purpose, active antennas that employ extended distributions of polarization currents moving faster than light have already demonstrated advantages over conventional technologies in communications and radar systems.

After a brief introduction, the present talk will focus on the mathematical treatment of extended faster-than-light sources of electromagnetic radiation, presenting ab-initio calculations and suggesting distributional as well as numerical solutions. This work was motivated by the fact that data collected from our practical emitters did not match the predictions made by others in the field; here we will carefully examine their arguments and rectify spurious claims that were made during the infancy of this new and vibrant field of research.

Photo: Derek Heeger

Derek Heeger [Fall 2020]
09-12-2020

Biography

Mr. Derek Heeger defended his PhD thesis on Thursday, November 12 at 2 pm in a Zoom meeting chaired by Dr. Jim Plusquellic. The title of Mr. Heeger's dissertation was "Mesh Networking and Security Applications in LoRa Ad-Hoc Networks."

Abstract

The Internet of Things (IoT) has rapidly expanded into consumer and industratial applications driving the need for secure and energy efficient communications suitable for battery powered devices. Common IoT protocols developed for these applications include SigFox, Narrow-Band IoT, and Long Range (LoRa). LoRa is an energy efficient long-range communication method enabled by advanced modulation schemes. This research enables new capabilities in LoRa ad-hoc networks allowing for innovative IoT sensors that were previously infeasible due to energy constraints. This is achieved by investigating adaptive data rate (ADR) techniques, enabling reliable and secure remote firmware updates, improving LoRa authenication, and implementing a secure mesh network topology suitable for mobile, energy-constrained devices. These concepts are developed, analyzed, and then applied to an advanced cattle monitoring sensor that tracks the health and location of free-range cattle.

Photo: Eric Hamke

Eric Hamke [Fall 2020]
09-14-2020

Biography

Mr. Eric Hamke defended his PhD thesis on Monday, September 14 at 11 am in a Zoom meeting chaired by Dr. Ramiro Jordan. The title of Mr. Hamke's dissertation was "Detecting Physical Stress Markers in Human Speech Using Deep Learning."

Abstract

The focus of the research has been to monitoring breathing and identify changes in pitch to identify physical task markers in speech.

Using a Restricted Boltzman Machine, firefighters' SCBA regulator valve sounds are classified as open and closed. This process also involves retraining periodically to include past observations since the last processing interval. The classifications are combined into continuous intervals. Observing the length of the intervals and the number of interval-starts represents time spent inhaling and the breathing rates (breaths per minute).

Voice pitch estimation using a multilayer LSTM network in a high noise envi-ronment (-10dB) uses a multilayer LSTM RNN classified speech segments as voiced or unvoiced. The classifications are grouped into continuous intervals. The pitch tracking processes uses these intervals in another multilayer LSTM RRN regression to estimate the pitch values and ensure continuity between observed values.

Photo: Mohamed Attia Abdelrahman Aref

Mohamed Attia Abdelrahman Aref [Fall 2020]
08-07-2020

Biography

Mohamed Attia Abdelrahman Aref defended his PhD thesis on Friday, August 7 at 10 am in a Zoom meeting chaired by Dr. Sudharman Jayaweera. The title of Mr. Aref's dissertation was "Cognitive Radios for Self-aware and Spectrum-agile Communications over Wideband Spectrum."

Abstract

In this dissertation, we introduce a novel cognitive engine design for intelligent, selfaware and spectrum-agile communications. The targeted cognitive radio (CR) has sensing and learning abilities that incorporate advanced signal processing and machine/deep learning (ML/DL) techniques. It can operate over a heterogeneous wideband spectrum that is much wider than the ones proposed in the literature, most likely covering several hundreds of MHz. It has the ability to detect existing signals in the surrounding RF environment and identify their origins. Furthermore, it cognitively uses channel selection, transmit power control and interference cancellation to adapt its communications mode inresponse to observed RF conditions. It can find the best transmission opportunities and maintain a certain level of quality of service (QoS). The proposed cognitive techniques can facilitate the use of intelligent spectrum management where operating licenses are issued based on the radio’s intelligence level. It also can be used in a wide range of applications including RF spectrum awareness, signals intelligence, anti-jamming and interference avoidance.

To this end, we implemented replicated Q-learning sub-band selection method for wideband spectrum based on partially-observable Markov decision process. Simulation results showed that the proposed algorithm can provide a substantial improvement over the random sub-band selection policy with only 16% below the optimal sub-band selection policy that requires complete state observability. A DL classification framework is introduced to distinguish between radar and communications signals based on their cyclostationary features. The proposed approach can reach classification accuracy around 99% at signal-to-noise ratio (SNR) of −4 dB. A multi-agent Q-learning technique is provided for anti-jamming and interference avoidance communications. An optimal energy detector is used based on the Neyman-Pearson (NP) criterion which maximizes the detection probability of existing signals subject to a certain false alarm level. Compared with existing schemes, the proposed approach uses two parallel communications operations: sensing and transmission, that can enable it to predict the locations of the jamming/interference signals accurately and hence avoid them efficiently.

A multi-task transfer deep reinforcement learning (DRL) approach is developed to provide spectrum-agile communications over wideband spectrum where communications over each sub-band represents a single task. The proposed approach attempts to learn an optimal policy for channel and transmit power level selection. The proposed approach has proved that it is suitable for real-time applications with its ability to adapt to sudden changes in the spectrum of interest. It outperforms existing schemes found in literature such as single-task DRL and Q-learning. A DL-aided successive interference cancellation (SIC) framework is proposed for massive multiple-input-multiple-output nonorthogonal multiple access (massive MIMO-NOMA) systems. A better realistic scenario, compared with similar approaches in the literature, that includes correlated channel fading and user mobility based on random waypoint (RWP) model is considered. Simulation results showed that the proposed approach provides a significant reduction in error propagation and computational complexity that might be encountered in traditional SIC schemes including those based on DL.

Photo: Luis Valbuena Reyes

Luis Valbuena Reyes [Fall 2020]
07-15-2020

Biography

Mr. Luis Valbuena Reyes defended his PhD thesis on Wednesday, July 15 at 2 pm in a Zoom meeting chaired by Dr. Edl Schamiloglu. The title of Mr. Reyes' dissertation was "Software Execution under Extreme Electromagnetic Interference."

Abstract

Software routines that deviate from the original algorithm goal due to a perturbation agent is a phenomenon that has been observed for a long time now. From unintended executions on flight control computers during the dawn of space exploration, passing through hardware Trojans tampering in corporate security, to sensitive medical equipment compromised to the point of changing the amount of delivered drug, the time intervals, and the reported biomedical measured data.

In this dissertation, a computer under perturbation is modeled by a hybrid system that captures the digital nature on the discrete part, and the interaction of the perturbation signal with the hardware on the continuous part, like extreme electro-magnetic interference (EEMI) on logic circuits. From this point, the dissertation develops in three directions. First, we develop a testbed simulation environment of a 4-bit processor with the ability to take perturbation signals both in time and space. Second, the formulation of mathematical models of the computer logic that take perturbation signals as inputs; and third, the use of reachability theory to access how far the perturbation propagates through hardware to cause an exploitable deviation in software execution. We formulate two modifications to the traditional Hamilton-Jacobi-Evans equations (HJE). A diffusion term is added to the original HJE and we also include the second-order terms of the Taylor expansion of the optimality principle.

Photo: Nishchay Sule

Nishchay Sule [Summer 2020]
06-30-2020

Biography

Nishchay Sule defended his PhD thesis on Tuesday, June 30 at 1 pm in a Zoom meeting chaired by Dr. Payman Zarkesh-Ha. The title of Mr. Esakki's dissertation was "Device-level Predictive Modeling of Extreme Electromagnetic Interference."

Abstract

Radio Frequency (RF) interference is a prominent issue for modern electronic devices. As device size and supply power shrink to meet the on-going demand for compact and complex Integrated Circuits (ICs), their susceptibility to external noise coupling to the input or power supply increases significantly. One such type of noise that acts upon a system to be considered is Extreme Electromagnetic Interference (EEMI). Previous works done to understand and evaluate the impact of EEMI onto a system or sub-system have been conducted on a statistical or empirical analysis level, which has led to complex and convoluted analysis, that requires significant time and computational power. Furthermore, since Electromagnetic Interference/Compatibility (EMI/EMC), engineers have to deal with complex systems, they typically come up with an estimate to analyze such systems. The premise behind this research is to highlight the development of the refinements of such "rule-of-thumb" guidelines to help EMI/EMC engineers quickly estimate device or circuit level susceptibility for the injected EEMI signals. A novel analytical model is proposed in this research, which offers an alternative solution for the limits of malfunction for a Silicon-based (Si) Metal-Oxide-Semiconductor Field-Effect Transistor (MOSFET) under EEMI bias.

These developed analytical predictive models help determine the maximum limits for large-signal gate-side (input) or drain-side (power supply) injection in terms of the device's ION/IOFF ratio prior to degradation or damage to the device. The ION/IOFF ratio is a function based on the MOSFETs' device parameters. These models have been developed for a single transistor, particularly n-type and p-type MOSFETs when the EEMI signal is superimposed onto the gate or drain terminals. Additionally, these predictive models have also been extended to determine the maximum tolerance for gate or drain injections as transistor technology scales down. Furthermore, these models have been compared and validated with prototype test chips across five different technology nodes.

Lastly, the analytical models have been expanded to be used in several different assessments, such as high-frequency analysis, manufacturer, and transistor size-independent and sensitivity modeling. Such demonstrations show the fundamental nature and flexibility, which allows these models to be used based on the EMI/EMC engineers' needs.

Photo: Shu Wang

Shu Wang [Summer 2020]
06-22-2020

Biography

Shu Wang defended his PhD thesis on Monday, June 22 at 1:30 pm in a Zoom meeting chaired by Dr. Christos Christodoulou. The title of Mr. Esakki's dissertation was "Advanced Parallel Algorithms in Computational Electromagnetics."

Abstract

The rapid development of high performance computing has pushed the computational electromagnetics (CEM) towards high accuracy, high fidelity and extreme computational scales. There is a great need for existing CEM solvers to have enhanced parallelism and scaling capability. The purpose of this dissertation is to investigate advanced parallel algorithms for both frequency and time domain solvers.

In frequency domain, this work first develop the underpinnings of parallel preconditioning technique and high-order transmission condition in the context of multi­solver scheme. The result is a computing resource-aware and implementation wise compact solver. Then this work targeted at developing efficient algorithms for cases where iteration of simulations,e.g. parameter sweep, is necessary. The proposed platform Green's function method can effectively reduce the turnaround time by exploiting reusable matrices.

In time domain, due to the ever-increasing sophistication in EM systems, a typical transient simulation may require many time steps. Most current transient solvers exploit parallelism in spatial domain which it is not trivial to sustain parallel scaling capability in high level. This work,therefore, provided a new perspective in parallelism, parallel-in-time(PIT). The problem is first decomposed based on superposition principle and corresponding effective integration methods are developed. Next, a hy­brid parallel scheme, space-time building block method, which is based on reduced order model, is proposed for applications like meta-material simulation. A improved scaling efficiency and 3x speed-up is observed in our work. Finally, PIT is extended to improve scaling efficiency for nonlinear circuit-electromagnetic co-simulations, where 2x better efficiency is achieved by proposed algorithms.

Photo: Mr. Gangadharan Esakki

Gangadharan Esakki [Summer 2020]
05-25-2020

Biography

Gangadharan Esakki defended his PhD thesis on Monday, May 25 at 9 am in a Zoom meeting chaired by Dr. Marios Pattichis. The title of Mr. Esakki's dissertation was "Adaptive Encoding for Constrained Video Delivery in HEVC, VP9, AV1 and VVC Compression Standards and Adaptation to Video Content."

Abstract

The dissertation proposes the use of a multi-objective optimization framework for designing and selecting among enhanced GOP configurations in video compression standards. The proposed methods achieve fine optimization over a set of general modes that include: (i) maximum video quality, (ii) minimum bitrate, (iii) maximum encoding rate (previously minimum encoding time mode) (iv) and can be shown to improve upon the YouTube/Netflix default encoder mode settings over a set of opposing constraints to guarantee satisfactory performance. The dissertation describes the implementation of a codec-agnostic approach using different video coding standards (x265, VP9, AV1) on a wide range of videos derived from different video datasets. The results demonstrate that the optimal encoding parameters obtained from the Pareto front space can provide significant bandwidth savings without sacrificing video quality. This is achieved by the use of effective regression models that allow for the selection of video encoding settings that are jointly optimal in the encoding time, bitrate, and video quality space. The dissertation applies the proposed methods to x265, VP9, AV1 and using new GOP configurations in x265, delivering over 40% of the optimal encodings in two standard reference videos. Then, the proposed encoding method is extended to use video content to determine constraints on video quality during real-time encoding. The content-based approach is demonstrated on identifying camera motions like panning, stationary and zooming in the video. Overall, the content-based approach gave bitrate savings of 35 % on the zooming & panning motion from Shields video, and 51.5 % on stationary & panning motion from Parkrun video. Additionally, the dissertation develops a segment-based encoding approach that delivers bitrate savings over YouTube’s recommended bitrates. Using BD-PSNR and BD-VMAF, a comparison is made of x265, VP9, AV1 against the emerging VVC encoding standard. The new VVC-VTM encoder is found to outperform all rival video codecs. Based on subjective video quality assessment study, AV1 was found to provide higher quality than x265 and VP9.

Photo: Jon Cameron Pouncey

Jon Cameron Pouncey [Spring 2020]
03-27-2020

Biography

Jon Cameron Pouncey defended his PhD thesis on Friday, March 27 at 10 am in room 118 of the ECE Building. Dr. Jane Lehr served as his committee chair. The title of Mr.Pouncey's dissertation is, "Enabling Compact Pulsed Power."  

Abstract

The first two decades of the 21st century have seen significant interest in expanding the application of pulsed power technology beyond traditional uses in physics and nuclear weapons research. Applications in the field of national defense, which present unique constraints on system size, have provided impetus to increase the exploration of compact pulsed power solutions. Innovations related to energy density, insulation, switching, and power conversion systems have been realized, bringing deployable compact pulsed power systems closer to realization than ever before. However, recent systems integration efforts have shown that research into tools and technologies is still needed to unlock the full potential of compact pulsed power. This thesis describes efforts to make contributions to three of these research needs.

The first need is for tools that improve predictive simulation of compact pulsed power systems. The principal challenge in satisfying this need is to develop simulation models of components unique to pulsed power systems for use in well-known simulation tools. In order to address this challenge, a predictive parametric model of a spark gap switch has been developed and validated for use in the ubiquitous SPICE circuit simulation software.

The second need is to formalize a set of design principles for compact pulsed power systems. The principal challenge in satisfying this need is to improve understanding of the behavior of pulsed power systems as they are made significantly more compact. This understanding can then be applied early in the development of systems to improve confidence that the system will not require significant redesign late in the development cycle. In order to address this challenge, a detailed analysis of the erection behavior of the compact Marx generator was completed, and translated into design improvement recommendations.

The third need is to leverage new technologies developed for commercial applications to advance compact pulsed power. The principal challenge in satisfying this need is to identify and assess technologies for applicability to compact pulsed power systems. In order to begin addressing this challenge, a series of novel experiments using a newly developed commercial infrared diode-pumped solid-state micro-laser to trigger gas switches were conducted.

Photo: Mitchell Martin

Mitchell Martin [Spring 2020]
04-08-2020

Biography

Mitchell Martin defended his PhD thesis on Wednesday, April 8 at 10 am in a Zoom video conference. Dr. James Plusquellic served as her committee chair. The title of Mr. Martin's dissertation is, "Physical Unclonable Functions Based on Delay Paths and an Interdigital Microstrip Notch Filter."  

Abstract

Physical Unclonable Functions Based on Delay Paths and an Interdigital Microstrip Notch Filter
Abstract: A physical unclonable function (PUF) is an integrated circuit hardware primitive that is designed to leverage naturally occurring variations to produce a random bitstring. The arbiter (ARB) PUF is one of the first to be described in the literature. It derives its entropy from variations that occur in the delays of identically configured logic paths. The ARB PUF uses a phase comparator to decide which path of a pair is faster under a given challenge and generates a 0 or 1 as a response indicator bit. Unfortunately, the ARB PUF is not reliable, requiring error correction in cases where the sequence of response bits (the bitstring) needs to be reproduced. In this proposal, a test structure is described, called a time-to-digital converter (TDC) that can measure the actual delays of the paths. This type of ’soft’ information can be used to improve the reliability of the ARB PUF. Data obtained from a set of chips fabricated in IBM’s 90 nm technology, and collected across 9 temperature-voltage corners, is used to demonstrate its effectiveness.

Current PUF designs are typically implemented in silicon like the ARB PUF or utilize variations found in commercial off-the-shelf (COTS) parts. Because of this, existing designs are insufficient for the authentication of Printed Circuit Boards (PCBs). In this thesis, we also propose a novel PUF design that leverages board variations in a manufactured PCB to generate unique and stable IDs for each PCB. In particular, a single copper trace is used as a source of randomness for bitstring generation. The trace connects three notch filter structures in series, each of which is designed to reject specific but separate frequencies. The bitstrings generated with both the ARB PUF and the PCB PUF are evaluated using statistical tests that measure randomness, uniqueness, and reliability.

Photo: Ran Luo

Ran Luo [Spring 2020]
03-16-2020

Biography

Ran Luo defended his PhD thesis on Monday, March 16 at 12:30 pm in room 118 of the ECE building. Dr. Yin Yang served as his committee chair. The title of Mr. Luo's dissertation is, "Advancing Elastic Solid Dynamics in Computer Graphics."  

Abstract

This dissertation proposes novel algorithms and applications and provides a real-time and easy-to-use simulator for realistic animation of the 3D solid model. The Finite Element Method (FEM) is a popular tool in the community because of its accurate result, however, the FEM is computationally expensive to handle a large number of DOFs. We present novel techniques to combine linear and nonlinear elasticity with model reduction to provide fast and realistic animation. On the other hand, one of the most important computation tasks of solid simulation is to evaluate the gradient vector and Hessian matrix of elastic energy function. We present a numerical routine to simplify the implementation of solid simulation with the complex-step finite difference (CSFD) that avoids subtractive cancellation. The complexity of nonlinearity is also an obstacle, and we provide a framework called NNWarp to combine the linear elasticity and neural network-based warping method to avoid expensive nonlinear optimization. We also propose an acoustic-VR system as the application. The system can convert acoustic signals of human language to realistic 3D tongue animation in real-time. The Deep Neural Networks (DNN) helps to convert the input speaking voice to positions of pre-defined EMA sensors. Then, a novel reduced physics-based solid simulator, introduced in previous, is used to synthesis the tongue animation.

 

Photo: Farhana Anwar

Farhana Anwar [Spring 2020]
03-10-2020

Biography

Farhana Anwar defended her PhD thesis on Tuesday, March 10 at 1 pm in the CHTM building. Dr. Ashwani Sharma served as her committee chair. The title of Ms. Anwar's dissertation is, "Tunneling and Transport Properties of Graphene."  

Abstract

This article provides a new method for computing electronic transport properties of graphene i.e. the peculiar tunneling properties of two-dimensional massless Dirac electrons. We consider a simple situation: a massless Dirac electron incident on a potential barrier which is tilted by applied bias and use finite difference method to obtain transmission probability(without involving transfer matrix). In the presence of an applied bias transmission coefficient and tunneling current were obtained and the effect of electric field which modulates the barrier profile therefore conductivity pattern were explained. Furthermore, this method can also be applied to investigate transport properties of disordered graphene as well as device characteristics of room temperature ballistic graphene field-effect transistors. Our study opens up a possibility for graphene-based device optimization by engineering barrier (gate) geometry for graphene MOSFETs as well as devices exploiting optics like behavior of Dirac fermions. In this work we show that finite difference method can be used as an effective approach in low dimensional semiconductor physics.

Photo: Joseph Gleason

Joseph Gleason [Fall 2019]
11-06-2019

Biography

Joseph Gleason defended his PhD thesis on Wednesday, November 6 at 4:30 pm in room 118 of the ECE building. Dr. Meeko Oishi served as his committee chair. The title of Mr. Gleason's dissertation is, "Software Design for Probabilistic Safety: Stochastic Reachability and Circadian Control."  

Abstract

Reachability is an important verification tool for that provides guarantees of safety while driving the system toward a target. For stochastic systems, in which there is some stochastic disturbance present in the dynamics, we seek controllers that provide a maximal likelihood of safety. Stochastic reachability analysis seeks to a) solves for controllers that maximize the likelihood that a system will reach a target state while staying in safe set, or b) determine the set of initial states for which there exists a controller that can reach a target while staying safe with at least a certain likeli­hood. Many domains are interested in solutions to stochastic reachability problems. An important example are biomedical systems. Because these systems interact with humans or are models that represent human biological functions, safety in control design is paramount. For biomedical systems, assurances of safety is a difficult be­cause many dynamical models of biological systems are often not well known and the health impacts of violations to safety constraints can have catastrophic consequences. Reachability analysis can provide a methodological framework that can work as the engineering equivalent of "do no harm".

This thesis covers topics in reachability and analysis of human circadian rhythms,i.e. the human sleep-wake cycle. We combine theoretical development with software implementation for both domains.

For stochastic reachability we propose methods for approximating stochastic reach­able sets using Lagrangian, set-based, methods. These allow for very fast computa­tion in low-dimensional systems when using polytopic sets, but provide conservative results. These methods utilize recursions that only use basic set operations. The theoretical development is applicable to nonlinear systems, but implementation is re­stricted to linear systems. We also describe SReachTools, a stochastic reachability toolbox which includes these Lagrangian methods as well as many other approximate and exact solutions for stochastic reachability problems.

In the biomedical domain, we examine methods for determining circadian phase in traumatically brain injured subjects and describe the implementation of a testbed designed to facilitate experimentation in circadian control. For assessing circadian rhythms we use continuously gathered biological signals to allow for a continuous assessment of circadian phase. We additionally demonstrate a unique correlation between circadian power and subject outcome. The lighting testbed details many design considerations for implementing a circadian control facility and demonstrates its ability to operate and its efficacy on phase advancing subjects with a simple pilot study. Both of these work toward an ultimate goal of being able to integrate assurances of safety when using feedback-based lighting to regulate human circadian cycles. This is an area of ongoing and future work.

Photo: Juan Briceno

Juan J. Faria-Briceno [Fall 2019]
11-04-2019

Biography

Juan J. Faria-Briceno defended his PhD thesis on Monday, November 4 at 9:30 am in room 101 of the CHTM building. Dr. Steven Brueck served as his committee chair. The title of Mr. Faria-Briceno's dissertation is, "Optical Angular Scatterometry: In-line Approach for Roll-2-roll and Nano-imprint Fabrication Systems."  

Abstract

As critical dimensions continue to shrink and structures become more complex, metrology processes are challenged to be implemented during in-line nanomanufacturing. Non-destructive, non-contact, and high-speed conditions are required to achieve proper metrology processes during in-line manufacturing. Optical scatterometry is a nanoscale metrology tool widely used in manufacturing. However, most applications of optical scatterometry operate off-line. A high-speed, in-line, non-contact, non-destructive scatterometry angular system has been demonstrated to scan pattern surfaces during real-time nano-fabrication.

Our system has demonstrated scanning capabilities using flat, 1D and 2D structures. The flat surface samples consist (commercially and grown) of thin film native oxide, grown oxide, and alumina. The 1D samples were made by using interferometric lithography with a thin deposited layer (~85 nm) of aluminum. The 2D complex samples are hollow silicon tubes fabricated using nano-imprint lithography. The inside diameter, outside diameter, and the period of the hollow pillars are respectively ~100 nm, ~135 nm and ~200 nm. These are test structures to establish the metrology capabilities. The applicability of the tool is not restricted to these samples.

Our current in-line scatterometer uses 45° off-axis parabolic mirrors which allows us to have an angular range up to ~50°. The system uses a high-speed scanner at 8 kHz to vary the angle of incident of the beam at the focal spot on the moving web. The system uses a 405 nm collimated diode laser. Our scan period is 125 µs which will allow us to scan 20-30 reflectance measurements before the web moves a distance comparable to the focal spot. A biased silicon detector is used to collect the high-speed reflection signal. The data collected is averaged with a digital scope and further processed on the computer for analysis. Our current system can be integrated with nano-imprint and other R2R real-time fabrication techniques. The goal is to improve quality control and monitor real-time high-speed nano fabrication processes. The angular range can be improved (up to 79°) by varying the focal length and the curvature of the parabolic mirrors. The scanning time can be reduced by increasing the frequency of the resonant scanner. Evidently, in-line angular scatterometry offers solutions to the future of R2R semiconductor nanomanufacturing.

 

Photo: Michael Darling

Michael Darling [Fall 2019]
11-01-2019

Biography

Michael Darling defended his PhD thesis on Friday, November 1 at 11 am in room 310 of the ECE building. Dr. Don Hush served as his committee chair. The title of Mr. Darling's dissertation is, "Using Uncertainty To Interpret Supervised Machine Learning Predictions."  

Abstract

When a machine learning model generates a prediction, a decision maker needs to determine its validity. Currently, such judgments depend heavily on expert opinion, using a mix of domain and modeling expertise, and accuracy-based validation
metrics. While these methods evaluate a model’s performance relative to a set of validation data, they do not tell us the model’s certainty with respect to a particular prediction on an unseen, perhaps critical, data point.

In this dissertation we develop an uncertainty measure we call minimum prediction deviation which we use to assess the quality of the individual predictions made by supervised two-class classifiers. We show how minimum prediction deviation can be used to di↵erentiate between the samples that a model predicts credibly, and the samples on which an alternate analysis may be required.

Photo: Vijay Saradhi Mangu

Vijay Saradhi Mangu [Fall 2019]
07-25-2019

Biography

Vijay Saradhi Mangu defended his PhD thesis on Thursday, July 25 at 9 am in room 101 of the CHTM building. Dr. Francesca Cavallo served as his committee chair. The title of Mr. Mangu's dissertation is, "Pixelated GaSb Membranes for Photovoltaics: Fabrication and Structure-Property Relationships."  

Abstract

The purpose of this thesis is to develop a reliable and efficient approach to heterogeneous integration of single-crystalline GaSb semiconductors with highly mismatched materials. The mismatch may refer to the crystalline structure and the thermal expansion coefficient of single-crystalline GaSb with respect to the other materials of interest. The strategy of hetero-integration relies on epitaxial lift-off (ELO). My approach prevents formation of extended structural defects that are detrimental to the performance of opto-electronic devices and preserves GaSb growth substrates for potential reuse.

Within my research work, I have overcome some outstanding challenges of epitaxial lift-off of lattice-matched GaSb epitaxial layers through pixelated approach and demonstrated the operation of single-crystalline GaSb photovoltaic devices with a unique architecture on single-crystalline Si substrates.

By leveraging release and transfer of GaSb membranes on Si, I have demonstrated operation of thin-film photovoltaic devices with areas of the of the order of 100sx100s mm2 (i.e., pixelated solar cells). The photo-conversion efficiency of ~ 340x340 mm2 pixelated devices amount to ~2.5%, i.e., a comparable efficiency to a 5 x 5 mm2 homo-epitaxial GaSb cell on a GaSb substrate.

A detailed structure-property relationships study is also performed to justify device characteristics of pixelated GaSb solar cells.

In conclusion, I have established a reliable and efficient process to isolate GaSb epilayers without the formation of any extended defects. I have demonstrated thin films and pixelated GaSb photovoltaic devices on single-crystalline Si substrates and performed a detailed structure-property relationship study to justify their properties and provide a path to improve their performance.

Photo: Alvaro Cerna

Alvaro Ulloa Cerna [Summer 2019]
07-03-2019

Biography

Alvaro Ulloa Cerna defended his PhD thesis on Wednesday, July 3 at 8 am in room 118 of the ECE building. Dr. Marios Pattichis served as his committee chair. The title of Mr. Cerna's dissertation is, "Large Scale Electronic Health Record Data and Echocardiography Video Analysis for Mortality Risk Prediction."  

Abstract

Electronic health records contain the clinical history of patients. The enormous potential for discovery in such a rich dataset is hampered by their complexity. We hypothesize that machine learning models trained on EHR data can predict future clinical events significantly better than current models. We analyze an EHR database of 594,862 Echocardiography studies from 272,280 unique patients with both unsupervised and supervised machine learning techniques.    

In the unsupervised approach, we first develop a simulation framework to evaluate a family of different clustering pipelines. We apply the optimized approach to 41,645 patients with heart failure without providing any survival information to the underlying clustering approach. The model separates patients with significantly different survival characteristics. For example, in a 10-cluster model, the minimum and maximum risk clusters had a median survival of 22 and 53 months respectively.

In the supervised approach, with 723,754 videos available from 27,028 unique patients, we assess the predictive capacity of Echocardiography video data for one- year mortality. Also, we hold out a balanced dataset of 600 patients to compare the model performance against cardiologists. We found that the best model, among four candidate architectures, is a 3D dyadic CNN model with an average AUC of 0.78 for a single parasternal long axis view. The model yields an accuracy of 75% (AUC of 0.8) on the held-out dataset while the cardiologists achieve 56% and 61%. The model performance was significantly higher than that of the cardiologists.

Finally, we develop a multi-modal supervised approach that enables interpretability. The model provides interpretations through polynomial transformations that describe the individual feature contribution and weights the transformed features to determine their importance. We validate our proposed approach using 31,278 videos from 26,793 patients. We test our proposed approach against logistic regression and non-linear and non-interpretable models based on Random Forests and XGBoost. Our results show that the proposed neural network architecture always outperforms logistic regression models while its performance approximates the other non-linear models. Overall, our multi-modal classifier based on 3D dyadic CNN and the interpretable neural network outperforms all other classifiers (AUC=0.83).

 

Photo: Saadat M. Mishkat-Ul-Masabih

Saadat M. Mishkat-Ul-Masabih [Summer 2019]
04-24-2019

Biography

Saadat M. Mishkat-Ul-Masabih defended his PhD thesis on Wed, April 24 at 10 am in room 101 of the CHTM Building. Dr. Daniel Feezell served as a committee chair. The title of Mr Saadat M. Mishkat-Ul-Masabih's dissertation is, "Nonpolar GaN-Based VCSELs with Lattice-Matched Nanoporous Distributed Bragg Reflector Mirrors."

Abstract

Wide-bandgap optoelectronic devices have undergone significant advancements with the advent of commercial light emitting diodes and edge-emitting lasers in the violet-blue spectral region. They are now ubiquitous in several lighting, communication, data storage, display, and sensing applications. Among the III-nitride emitters, vertical-cavity surface-emitting lasers (VCSELs) have attracted significant attention in recent years due to their inherent advantages over edge-emitting lasers. The small active volume enables single-mode operation with low threshold currents and high modulation bandwidths. Their surface-normal device geometry is conducive to the cost-effective formation of high-density 2D arrays while simplifying on-chip wafer testing. Furthermore, the low beam divergence and circular beam profiles in VCSELs allow efficient fiber coupling.

Nevertheless, GaN-based VCSELs are still in the early stages of development. Several challenges need to be addressed before high-performance devices can be commercially realized. One such challenge is the lack of high-quality distributed Bragg reflector (DBR) mirrors. Conventionally, epitaxial and dielectric DBRs are used which often involve complex growth and fabrication techniques. This dissertation provides an alternative approach where subwavelength air-voids (nanopores) are introduced in alternating layers of doped/undoped GaN to form the DBR structure. Selective electrochemical etching creates nanopores in the doped layers, reducing the effective refractive index relative to the surrounding undoped GaN. Using only 16-pairs, DBR reflectance >99.9% could be achieved. Several research groups have shown optically pumped VCSELs using nanoporous DBRs on c-plane. However, there are no reports of electrically injected nanoporous VCSELs. Using m-plane GaN substrates, we have demonstrated the first ever electrically injected GaN-based VCSEL using a lattice-matched nanoporous DBR. The nonpolar m-plane orientation is beneficial for leveraging the higher per-pass gain and polarization-pinning properties absent in c-plane. Lasing under pulsed operation at room temperature was observed at 409 nm with a linewidth of ~0.6 nm and a maximum output power of ~1.5 mW. This is the highest output power from m-plane VCSELs to date with relatively stable operation at elevated temperatures. All tested devices were linearly polarization-pinned in the a-direction with high polarization ratios >0.9. Overall, the nanoporous DBRs help in mitigating some of the issues that limit the performance of III-nitride VCSELs.
Photo: Brock Roberts

Brock Roberts [Spring 2019]
04-15-2019

Biography

Brock Roberts defended his PhD thesis on Monday, April 15, 2019 at 1 pm in room 237 of the ECE Building. Dr. Edl Schamiloglu served as his committee chair. The title of Mr. Robert's dissertation is, "Noise and Gain Characterization of Interband Cascade Infrared Photodetectors."

Abstract

A cavity designed to have multiple harmonic TM0N0 modes can be used to accurately measure the longitudinal profile of a bunched charged particle beam passing through its bore, non-invasively, and in real time. 

Multi-harmonic TM0N0 cavities were designed, constructed, and beamline tested in a variety of experiments at the Thomas Jefferson National Accelerator Facility (TJNAF or Jlab).   Measurements with a sampling oscilloscope provided signals that resemble the profile of electron bunches passing through the cavity’s bore.  Straightforward signal processing techniques reduce distortion in the measurement and provide real time profiles of electron bunches with picosecond accuracy.  Subharmonic beams having bunch repetition rates of 1/3rd and 1/6th of Jlab’s 1497 MHz bunch frequency, and interleaved sub-harmonic beams were also measured.  Comparison between measurements made using a harmonic cavity were corroborated with an established invasive measurement method and with computer models.   A harmonic cavity from this effort has been installed within the CEBAF injector, allowing accelerator operators to view, in real time, the shape and duration of electron bunches entering the accelerator.  Another harmonic cavity has been installed within Jlab’s Upgraded Injector Test Facility (UITF), and two more are planned for installation there.   This effort was awarded the 2016 International Beam Instrumentation Conference’s Faraday Cup Award.


 

Photo: Francisco German Perez Venegas

Francisco German Perez Venegas [Spring 2019]
04-12-2019

Biography

Francisco German Perez Venegas defended his PhD thesis on Friday, April 12 at 2:30 pm in room 118 of the ECE Building. Dr. Balu Santhanam served as his committee chair. The title of Mr. Venegas' dissertation is, "Detection and classification of vibrating objects in SAR images."

Abstract

The vibratory response of buildings and machines contains key information that can be exploited to infer their operating conditions and to diagnose failures. Further­more, since vibration signatures observed from the exterior surfaces of structures are intrinsically linked to the type of machinery operating inside of them, the ability to monitor vibrations remotely can enable the detection and identification of concealed machinery.

This dissertation focuses on developing novel detection schemes for performing detection and classification of vibrating objects in SAR images. Three are the cen­tral claims of this dissertation. First, the non-linear transformation that the micro­doppler return of a vibrating object suffers through SAR sensing does not destroy its information. Second, the instantaneous frequency (IF) of the SAR signal has sufficient information to characterize vibrating objects. Third, it is possible to de­velop a detection model that encompasses multiple scenarios including both mono­component and multi-component vibrating objects immersed in noise and clutter.

For answering these claims, two different schemes are studied for both the de­tection and classification of vibrating objects in SAR images. The first scheme is data-driven and utilizes features extracted with the discrete fractional Fourier transform (DFRF T) to feed machine-learning algorithms (MLAs). Specifically, the DFRFT is applied to the IF of the slow-time SAR data, which is reconstructed using techniques of time-frequency analysis. The second scheme is model-based and employs a probabilistic model of the SAR slow-time signal, the Karhunen-Loeve transform (KLT), and a likelihood-based decision function. The performance of the proposed schemes is characterized using simulated data as well as real SAR data co­llected with the Lynx SAR. The suitability of SAR for sensing vibrations is demons­trated by showing that the separability of different classes of vibrating objects is preserved even after non-linear SAR processing.

Finally, the proposed algorithms are studied in the presence of signal noise and terrain clutter. The results show that the proposed schemes produce high-precision classifiers capable of dealing with noise and clutter of moderate intensity. In order to loosen these requirements, the Hankel rank reduction (HRR) method, previously used for suppressing ocean clutter in ground-wave radar, is adapted to suppress clutter-noise in SAR images.

Photo: Joshua J. Trujillo, Sr.

Joshua J. Trujillo, Sr. [Fall 2019]
04-10-2019

Biography

Joshua J. Trujillo, Sr. defended his PhD thesis on Wednesday, April 10 at 11:30 am in room 118 of the ECE building. Dr. Payman Zarkesh-Ha served as his committee chair. The title of Mr. Trujillo's dissertation is, "Study of Distributed Versus Compressed Layouts for PUFs."  

Abstract

Society continues to depend on electronics for everything from smart systems in our homes to cellphones and tablets, which are more powerful today than many desktop computers that are still in use [1]. With increased consumption of these electronic products comes an increase in problems, such as counterfeit integrated circuits being sold as genuine integrated circuits. This is just one of many problems that corporations and end users are having to deal with in this digital age.

Software security has been accepted as a problem by both the media and general public, but only recently has hardware security begun to come out as a problem as important, if not more important, than software security. When there is an issue with software, a software patch can be released to fix this issue. Due to the nature of hardware, this is not possible with hardware security problems. Often times, frrmware can be updated to partially address the problem, as in the case with the Intel and AMD processor flaws in the news recently [2]. However, there are many times that the only solution is to take the hardware system offline and replace the questionable components.

Hardware security researchers have been trying to fmd ways to use security circuits in hardware design to help combat this problem. One type of security circuit being researched and used today is called a Physically Unclonable Circuit (PUF). PUF circuits can be added onto an existing integrated circuit design, interrogated at fabrication for challenge-response pairs which are stored in a database, then checked against that database at any time in the lifecycle of that integrated circuit [3].

This research contributes to PUF security circuits by showing how different ways of laying out a circuit design on an integrated circuit can change the performance of the circuit, as well as if PUF circuits designed on SiGe BiCMOS can be used in the same way that PUF circuits fabricated using standard CMOS processes are used today.

Photo: Lilian Casias

Lilian Casias [Spring 2019]
03-20-2019

Biography

Lilian Casias defended her PhD thesis on Wednesday, March 20 at 10 am at the CHTM building. Dr. Ganesh Balakrishnan served as her committee chair. The title of Ms. Casias' dissertation is, "Transport in Mid-Wavelength Infrared (MWIR) p- and n- type InAsSb and InAs/InAsSb Type-II Strained Layer Superlattices (T2SLs) for infrared detection."

Abstract

III-V materials such as InAsSb ternaries and InAs/InAsSb Type-II Strained Layer Superlattices (T2SLs) have significant potential for infrared (IR) detector applications, including space-based detection, when utilized in a unipolar barrier detector architecture (nBn). However, recent studies revealed the quantum efficiency in nBn detectors degrades significantly faster from proton-irradiation induced displacement damage as compared to HgCdTe photodiodes. Improving the quantum efficiency radiation-tolerance is theoretically possible by enhancing vertical hole mobility and thereby the vertical hole diffusion length. The vertical hole mobility of T2SLs materials differs significantly from the lateral mobility and measuring it is much less straightforward.

In order to tackle vertical transport, in-plane or lateral transport must be better understood. There are added challenges to determining the in-plane bulk carrier concentration in narrow bandgap materials due to the potential for electron accumulation at the surface of the material and at its interface with the layer grown directly below it. Electron accumulation layers form high conductance electron channels that can dominate both resistivity and Hall-effect transport measurements. Therefore, to correctly determine the in-plane bulk concentration and mobility, temperature- and magnetic-field-dependent transport measurements in conjunction with Multi-Carrier Fit (MCF) analysis were utilized on a series of p-doped InAs0.91Sb0.09 samples on GaSb substrates. The samples are etched to different thicknesses and variable-field measurements are utilized to assist in confirming whether a carrier species represents bulk, interface or surface conduction.

Secondly, n-type temperature- and magnetic-field dependent measurements on InAsSb and InAs/InAsSb T2SLs materials were performed to extract the in-plane transport properties for all the carriers present in each sample under two different doping concentrations (undoped and Silicon-doped). Lastly, substrate-removed, metal-semiconductor-metal (MSM) devices were fabricated to attempt vertical measurements, while standard van der Pauw structures were used for in-plane measurements. The MSM processing serves as a potential fabrication technique to measure vertical transport, that can be improved in the future. The goal of this dissertation is to accurately determine the lateral and vertical transport properties in the presence of multiple carrier species, Multi-Carrier Fit (MCF) and High-Resolution Mobility Spectrum Analysis (HR-MSA) were employed.



Photo: Eli Garduno

Eli Garduno [Spring 2019]
09-06-2018

Biography

Eli Garduno defended his PhD thesis on Thursday Sept. 6, 2018 at 10 am in the CHTM Building. Dr. Ganesh Balakrishnan served as his committee chair. The title of Mr. Garduno's dissertation is, "Noise and Gain Characterization of Interband Cascade Infrared Photodetectors."

Abstract

Infrared (IR) detectors are an enabling technology for a broad and growing list of applications including gas detection, night vision, and space-based missile warning. There are ongoing efforts in IR detector research to explore the potential of new material systems and energy band structures in addition to continuously improving their sensitivity through increasing their quantum efficiency and lowering their dark current and noise. This dissertation examines an emerging class of IR detectors known as Interband Cascade Infrared Photodetectors (ICIPs).

ICIPs contain multiple regions to facilitate the collection of photogenerated elec­trons and to limit unwanted dark current. Theory regarding their performance also indicates that multi-stage ICIPs may have lower noise than single-stage ICIPs and may provide improved detectivity in cases where the absorption coefficient of a ma­terial system is small and/ or where the diffusion length in the material is short or degraded.

In this work, four long-wavelength infrared ICIP devices with one, four, six, and eight stages were characterized at varying temperatures from 80 to 300 K and at biases up to one volt in both forward and reverse polarities. Noise spectra were collected on the four devices and show significant 1 / f noise that prevented direct measurement of the ICIP noise gain. The 1/ f noise in the ICIPs was linked to generation-recombination current. The devices were found to cause circuit instability when operated in bias regions with negative differential conductance (NDC) due to bias-dependent resonant tunneling. Additionally, bias-dependent photocurrent gain was observed using illumination of the devices with 632 nm and 1550 nm lasers which peaked near the NDC regions. This photocurrent gain was experimentally shown to be caused by current-mismatch between device stages, verifying theories regarding its origin.

Photo: Nicholas Tarasenko

Nicholas Tarasenko [Spring 2019]
02-21-2018

Biography

Nicholas Tarasenko defended his PhD thesis on Thursday, February 21, 2019 at 1 pm in room 118 of the ECE Building. Dr. Christos Christodoulou served as his committee chair. The title of Mr. Tarasenko's dissertation is, "DESIGN AND IMPLEMENTATION OF A 72 & 84 GHZ TERRESTRIAL PROPAGATION EXPERIMENT; EXPLOITATION OF NEXRAD DATA TO STATISTICALLY ESTIMATE RAIN ATTENUATION AT 72 GHZ."

Abstract

The wireless communication sector is rapidly approaching network capacities as a direct result of increasing mobile broadband data demands. In response, the Federal Communications Commission allocated 71-76 GHz “V-band” and 81-86 GHz “W-band” for terrestrial and satellite broadcasting services. Movement by the telecommunication industry towards W/V-band operations is encumbered by a lack of validated and verified propagation models, specifically models to predict attenuation due to rain. Additionally, there is insufficient data available at W/V-bands to develop or test propagation models. The first aim of this study was the successful installation and operation of a terrestrial link to collect propagation data at W/V-band frequencies. In September 2015, the University of New Mexico, in collaboration with the Air Force Research Laboratory’s Space Vehicle Directorate, NASA’s Glenn Research Center and industry partners including (ACME, Applied Technology Associates, and Quinstar Technologies, Inc.) established the W/V-band Terrestrial Link Experiment (WTLE). WTLE was installed in the Albuquerque metro area with persistent tonal transmissions at 72 GHz and 84 GHz on a 23.5 km slanted path.

The second aim of this study was the utilization of the National Weather Service’s Next Generation Weather Radar (NEXRAD) system data to statistically estimate attenuation due to rain at 72 GHz. NEXRAD data provides a distributed sense of rain rates along WTLE’s path and alleviates challenges associated with instrumenting the 23.5 km link. Furthermore, NEXRAD data alleviates the need to develop complicated routines using in-situ meteorological measurements to estimate the size of the rain cell affecting the link. Non-linear regression techniques were applied on 2017 monsoon season data to obtain rain rate power law model coefficients. Testing of these coefficients was conducted on 2018 monsoon season data with satisfactory results. The techniques employed in this analysis represent a significant advancement in the ability to predict attenuation due to rain at 72 GHz for terrestrial links by enabling the use of historical archives of publicly available National Weather Service NEXRAD data. The technique has promising potential for estimation of path attenuation due to rain for links other than WTLE because of the vast nationwide coverage provided by NEXRAD systems.