Electrical and Computer Engineering

Permanent URI for this collectionhttps://uwspace.uwaterloo.ca/handle/10012/9908

This is the collection for the University of Waterloo's Department of Electrical and Computer Engineering.

Research outputs are organized by type (eg. Master Thesis, Article, Conference Paper).

Waterloo faculty, students, and staff can contact us or visit the UWSpace guide to learn more about depositing their research.

Browse

Now showing 1 - 20 of 2143

Modelling of a Small Electric Aircraft Pipistrel Velis Electro
(University of Waterloo, 2026-01-21) DASARI MURUGAPPA, LEKHA
Transportation electrification has become an active area of research and development in academia and industry, with a strong focus on decarbonizing the sector to move toward a more sustainable environment. As an important player in the global sustainable transportation movement, the aviation industry is also witnessing accelerated efforts towards electrification. This transition comes with many challenges in terms of battery performance, aircraft flight range, and operational safety. Therefore, development of comprehensive simulation models, that replicate the behavior of an actual aircraft, is essential for studying the system’s overall performance. Such models provide invaluable insights into battery health, methods to extend range, and ways to improve flight missions for more efficient battery usage. This thesis aims to develop a mathematical model of the aircraft propeller and a simulation model of the electric powertrain consisting of the battery pack, inverter, and motor. The aircraft under study is the Pipistrel Velis Electro, a two-seater, type-certified, fully electric aircraft. Two methods are proposed to model the propeller behavior: one based on the aircraft equations of motion and the other based on the motor power command. Both methods compute the thrust, motor rotational speed, and load torque for each phase of the flight using different input sets, and these outputs are supplied to the powertrain model. Two modes of operation are considered for the powertrain: an autopilot flight mode and a pilot-controlled mode, with phase detection between powered and glide phases. The simulations are validated by comparing the results with actual Velis Electro flight data obtained from the Waterloo Wellington Flight Center.
JPEG-Inspired Encoding for Deep Learning
(University of Waterloo, 2026-01-21) Salamah, Ahmed Hussein Abdallah Mohamed
JPEG is the dominant standard for storing and transmitting digital images, while Deep Neural Networks (DNNs) have become the preeminent method for automated image understanding. This dissertation investigates how these two ubiquitous technologies can be synergistically integrated to enhance the performance of DNNs. JPEG was originally engineered for the Human Visual System (HVS), and its default parameters are not optimized for DNNs, which process visual information differently. This suboptimality, stemming from JPEG’s default implementation, is not a fundamental limitation but rather an opportunity to adapt its core components—especially the non-linear quantization stage—for DNNs. This research addresses this suboptimality by first optimizing the trade-off between compression rate and classification accuracy, and second, by introducing a learnable, end-to-end differentiable JPEG layer whose quantization parameters are jointly trained with the underlying DNN. This dissertation demonstrates that this principle of a learnable, JPEG-inspired transformation extends beyond compression, offering a novel way to address challenges in related domains such as knowledge distillation (KD), where large 'teacher' models often overfit the training set. This overfitting causes them to generate overconfident, near one-hot probability vectors that serve as poor supervisory signals for the student model, suggesting the need for novel approaches to information transfer. This dissertation addresses these issues by systematically revisiting the relationship between JPEG encoding and deep learning. It charts a logical progression from adapting JPEG externally for DNNs, to integrating it internally as a learnable network component, and finally to repurposing its core principles to amplify knowledge transfer. This progressive framework is methodically developed and empirically substantiated through three interconnected contributions: -Optimizing Compression for DNNs. To improve the interaction between standard JPEG and pre-trained DNNs, this work first reframes compression from a human-centric "rate-distortion" problem to a DNN-centric "rate-accuracy" one. This is achieved by introducing the Sensitivity Weighted Error (SWE), a novel distortion measure derived from a DNN’s loss sensitivity to frequency-domain perturbations, where higher sensitivity in a frequency band indicates its greater importance for the DNN’s decision-making. The SWE guides the OptS algorithm to generate model-specific JPEG quantization tables. This approach produces fully compliant JPEGs optimized for DNN consumption, demonstrably improving the rate-accuracy trade-off by increasing accuracy up to 2.12% at the same rate, or enabling rate reductions up to 67.84% with no loss of model accuracy. -Integrating a Differentiable JPEG Layer into the DNN Architecture. Building on this, the next contribution integrates the codec into the network architecture itself via the JPEG-Inspired Deep Learning (JPEG-DL) framework, which introduces a novel, end-to-end differentiable JPEG layer. By replacing JPEG's standard hard quantization with a differentiable alternative, this layer's parameters are jointly optimized with the network's weights. This transforms the JPEG pipeline from a static pre-processor into a dynamic, learnable component, significantly improving model accuracy—by an average of 7% on fine-grained classification tasks with only 128 additional trainable parameters—and enhancing robustness against adversarial attacks. - Amplifying Knowledge Transfer via JPEG-Inspired Perturbation. Finally, the differentiable layer is repurposed to address the "overconfident teacher" problem in KD by perturbing teacher inputs to force softer, more informative predictions. Crucially, this method requires no retraining or modification of the fixed teacher model, ensuring its practical utility with proprietary or deployed networks. Our investigation begins with Coded Knowledge Distillation (CKD), a practical heuristic that uses adaptive JPEG compression to perturb teacher inputs and soften their overconfident predictions. While effective, this approach prompted a search for a more principled theoretical foundation. This led to Generalized Coded Knowledge Distillation (GCKD), a framework that establishes the maximization of the teacher's Conditional Mutual Information (CMI) as the core objective. However, directly optimizing for CMI on a per-input basis is computationally prohibitive. This efficiency challenge is resolved in the culminating synthesis, Differentiable JPEG-based Input Perturbation (DJIP). DJIP operationalizes the GCKD theory by deploying the trainable differentiable JPEG layer as a fast, learnable, and amortized operator. Instead of performing a slow, per-input optimization search, the layer is trained once to automatically generate CMI-maximizing perturbations, making the process highly efficient. This approach demonstrably generates richer supervisory signals, boosting student model accuracy by up to 4.11%. In conclusion, this dissertation demonstrates that the relationship between JPEG and DNNs can be systematically revisited to create a powerful synergy. By progressing from adaptation to integration and synthesis, this work transforms the suboptimal default interaction of JPEG and DNNs into a versatile architectural tool. The research delivers a suite of methods that not only improve the performance of DNNs on compressed images but also offer a theoretically-grounded solution to a key challenge in knowledge distillation. By demonstrating that legacy codecs can be repurposed to enhance model accuracy, efficiency, and knowledge transfer, this work thus reframes the role of classical codecs, proposing JPEG-inspired encoding as a principled foundation for the integration of classical compression and deep learning.
Computer-Aided Modeling and Tuning of RF Acoustic Wave Filters
(University of Waterloo, 2025-12-19) Ou, Matthew
Mobile radios must operate across many frequency bands while sharing antennas and supporting closely spaced transmit and receive paths. Achieving this requires high-performance RF filtering that suppresses interference, limits unwanted emissions, and provides strong isolation. In smartphones, these functions are predominantly realized using acoustic-wave devices, particularly SAW- and BAW-based resonator filters implemented as duplexers and multiplexers. Their high selectivity, low insertion loss, and compact size have displaced conventional solutions as systems expanded from limited third-generation band sets to much larger fifth-generation portfolios. This growth in band count has increased both the number and diversity of filters within a single platform, making acoustic filters a dominant contributor to RF front-end cost and area. This dissertation addresses key challenges in the dynamic tuning of acoustic filters and in late-stage band targeting under process-induced detuning. Several bandwidth-reconfigurable architectures are introduced that integrate switches with resonators to enable bandwidth reconfiguration. In addition, the dissertation demonstrates that strategic modification of interdigital transducer configurations in ladder acoustic filters enables discrete control of the resonator electromechanical coupling coefficient. By adjusting the ratio of positive to negative IDT fingers, coupling can be set to selected levels to improve selectivity and asymmetry and to meet diverse specifications. VO₂-based RF switches are integrated directly with SAW resonators to select discrete IDT states, enabling ladder bandwidth tuning while maintaining low insertion loss. The dissertation further demonstrates a computer-aided tuning framework for late-stage band targeting in acoustic filters. A resonator-level extraction method reconstructs ladder element parameters directly from measured filter responses using stable rational approximation of the driving-point function, pole-zero identification via partial and continued fraction expansions, topology-aligned element matching, and sequential decomposition of series and shunt resonators. In addition, an on-wafer tuning strategy is demonstrated that prescribes minimal, physically realizable corrections without embedded tuning components or manual intervention. The approach spatially programs mass loading through patterned dielectric overlays for additive shifts and electrode thinning or ion milling for subtractive shifts, enabling heterogeneous per-resonator trims simultaneously. Experimental results demonstrate recovery of target passband characteristics, with improved return loss and insertion loss, establishing a practical framework for acoustic-filter correction under manufacturing-induced non-idealities.
Investigation of Organic ETLs in QLEDs and a Metal-based RGB Patterning Technique for QLED Displays
(University of Waterloo, 2025-12-17) Mobarak, Saad
Colloidal quantum dot light-emitting devices (QLEDs) have attracted significant interest for next-generation emissive display and lighting applications owing to their narrowband emission, tunable bandgaps, and compatibility with solution-based fabrication. Their emissive layers (EMLs), composed of colloidal quantum dots (QDs), exhibit discrete energy states and size-dependent bandgaps, allowing precise spectral tunability and narrow emission linewidths (FWHM < 25 nm). The high photoluminescence quantum yield (PLQY), excellent photochemical stability, and compatibility with low-temperature fabrication processes make them highly suitable for large-area and flexible devices. Collectively, these properties position QLEDs as strong contenders to replace organic-LEDs (OLEDs) in future display technologies, offering improved color saturation, reduced power consumption, and enhanced manufacturing versatility. Despite these advantages, QLEDs still face fundamental challenges related to charge transport and device efficiency. In particular, the use of organic electron transport layers (ETLs) has been limited due to their perceived low electron mobility and inferior performance compared to inorganic metal-oxide ETLs. However, organic ETLs remain attractive for certain device architectures because of their solution processability, tunable energy levels, and compatibility with low-temperature and flexible fabrication. Moreover, organic layers can form smoother, defect-free interfaces with the QD EMLs compared to metal-oxide ETLs, which may introduce interfacial traps or cause damage during deposition. While the low efficiency of QLEDs employing organic ETLs has conventionally been attributed to their poor electron mobility, the findings presented in this thesis reveal that uncontrolled electron leakage from the QD EML to the hole transport layer (HTL) plays a more dominant role. Based on the finding, the design and optimization of multilayer organic ETL architectures with electron-blocking interfaces effectively suppress electron leakage, leading to improved charge balance and enhanced device efficiency. Using this approach, both red and green QLEDs achieve maximum EQEs approaching 10%, representing among the highest reported values for devices employing organic ETLs. Another limitation in QLEDs is their limited amenability to high-resolution patterning of RGB arrays for full-color displays. Conventional techniques, such as inkjet printing or photolithography, often suffer from limited resolution, QD degradation, or complex processing steps that can compromise device performance. This thesis also presents a novel RGB patterning technique based on metal-induced quenching. Thin metal layers are selectively deposited via a shadow mask onto target areas of the QD EMLs, where subsequent metal diffusion into the EML locally suppresses luminescence through non-radiative energy transfer, while unexposed regions retain their intrinsic emission characteristics. Optical and morphological characterization shows that metal-coated QD regions develop increased surface roughness and island-like features, indicating that metal diffusion into the QD layer plays a significant role in facilitating non-radiative quenching. Using this approach, we demonstrate the fabrication of devices containing multiple QLEDs from a single multilayer stack, each producing spectrally pure electroluminescence (EL) without detectable parasitic emission. Additional patterned structure demonstrates controlled microscale emission at the device level, establishing the feasibility of achieving spatial color definition with high precision. These results validate metal-induced quenching as an effective methodology for QLED color patterning and provide insight into metal-QD interactions.
Multiple Model Adaptive Control with Blending in State Space
(University of Waterloo, 2025-12-11) Lovi, Alex
Adaptive Control (AC) provides a systematic framework for handling uncertainty in linear and non-linear systems. Single-model adaptive schemes such as Model Reference Adaptive Control (MRAC) and Adaptive Pole Placement Control (APPC) face inherent limitations when applied to systems with large parametric uncertainty, such as slow convergence rates and limited noise robustness. This has motivated researchers to investigate multiple-model strategies that employ several candidate plants to represent different regions of the operating space. In this thesis, we develop Multiple-Model Adaptive Control (MMAC) methodologies based on the blending of signals from multiple fixed models. We consider uncertain plants with known, compact, convex polytopic uncertainty. Our starting point is the design of a Multiple-Model Parameter Identification (MMPI) scheme that quickly and robustly identifies the uncertain plant parameters. In combination with a Model Reference Control (MRC) framework, this leads to a Multiple-Model Reference Adaptive Control (MMRAC) with blending for Linear, Time-Invariant (LTI), non-square (different number of inputs and outputs), multi-input systems, with full-state feedback. Under an exact matching condition, the parameter estimates are used to design a control input such that the plant states asymptotically track the reference signal generated by a state-space reference model. A procedure is provided to select the corner models based solely on the polytopic uncertainty. The proposed MMRAC guarantees the boundedness of all closed-loop signals and the asymptotic convergence of the state-tracking error to zero. Statistical analysis demonstrate improved tracking speed and robustness to noise compared with single-model approaches. The combination of MMPI with pole-placement techniques, allowed us to develop Multiple-Model Adaptive Pole Placement Control (MMAPPC) for LTI, square (same number of inputs and outputs), multivariable systems with full-state feedback, and for with LTI, Single-Input, Single-Output (SISO) systems via an intermediate state estimation step. The resulting controller again ensures the boundedness of all closed-loop signals, while also asymptotically placing the closed-loop eigenvalues at designer-specified locations. Statistical analysis shows a clear increase in robustness to noise relative to single-model schemes. These improvements were validated in the context of motion control of lateral vehicle dynamics, where multiple-model schemes consistently outperformed single-model approaches, including cases with slowly time-varying unknown parameters.
Algorithmic Collective Action under Differential Privacy
(University of Waterloo, 2025-12-11) Solanki, Rushabh
The rapid integration of AI into everyday life has generated considerable public attention and excitement, pointing to unprecedented gains in efficiency and personalization. However, it concurrently raises concerns about the potential for automating algorithmic harms, as well as re-entrenching existing social inequities. In response, the pursuit of "trustworthy AI" has become a critical goal for researchers, corporations, and governments alike. Achieving this objective is a complex challenge with many possible ways to realize it, ranging from policy and regulation to technical solutions such as algorithmic design, systematic evaluation, and enhanced model transparency. Contemporary AI systems, particularly those leveraging large-scale models, are fundamentally trained on vast amounts of data, often sourced from social interactions and user-generated content. This dependence introduces a grassroots mechanism for autonomy: the possibility for everyday users, citizens, or workers to directly steer AI behavior. This concept, known as Algorithmic Collective Action (ACA), involves a coordinated effort where a group of individuals deliberately modifies the data they share with a platform, with the intention of driving the model’s learning process toward outcomes they regard as more favorable or equitable. We investigate the intersection between these bottom-up, user-driven efforts to influence AI and a growing class of methods that firms already implement to improve model trustworthiness, especially privacy protections. In particular, we focus on the setting in which an AI firm deploys a differentially private model, motivated by the growing regulatory focus on privacy and data protection. Differential Privacy (DP) is a formal, mathematical framework that provides provable guarantees about the privacy of individuals whose data are used in a dataset. To operationalize these privacy guarantees in deep learning settings, we employ Differentially Private Stochastic Gradient Descent (DP-SGD), which is the de facto DP mechanism for deep learning, making it a natural choice for assessing the effectiveness of ACA under realistic conditions. Our findings reveal that while differential privacy offers substantive protection for individual data, it concurrently introduces challenges for effective algorithmic collective action. Theoretically, we characterize formal lower bounds on the success of ACA when a model is trained with differential privacy. These bounds are expressed as a function of key variables: the size of the acting collective and the firm’s chosen privacy parameters, which dictate the level of privacy the firm intends to enforce. Empirically, we verify these theoretical trends through extensive experimentation by simulating collective action during the training of a deep neural network classifier across several datasets. Moreover, we offer additional insight into how ACA affects empirical privacy, and we include a socio-technical discussion of the wider implications for responsible AI.
Wideband Reconfigurable Intelligent Surfaces
(University of Waterloo, 2025-12-10) Tayebpour, Jalaledin
Reconfigurable intelligent surfaces (RISs) have emerged as one of the most significant innovations in wireless communications, offering a novel approach to meeting the escalating demand for higher data rates, seamless coverage, and energy-efficient connectivity in next-generation networks. Unlike conventional wireless systems that rely on active and power-hungry components, RIS leverages nearly passive reflecting elements arranged in a planar array, whose electromagnetic responses can be dynamically reconfigured. By enabling programmable control over the incident wavefront, RIS introduces a new paradigm in which the wireless environment itself becomes a controllable entity. This capability not only enhances system performance but also reduces overall power consumption and hardware complexity, positioning RIS as a key enabler for sixth-generation (6G) and beyond communication technologies. This thesis provides a comprehensive investigation into the principles, design methodologies, and system-level benefits of RIS technology. The research begins with an in-depth review of the current state of the art, highlighting both theoretical foundations and practical implementations of RIS. Building on this foundation, the work develops novel design strategies for reconfigurable unit cells intended for RIS applications. Several geometries are explored with the goal of achieving tunable reflection phase profiles, wide operational bandwidth, and multi-polarization capability. The designs integrate semiconductor switches and micro-electromechanical systems (MEMS) actuators, demonstrating the feasibility of programmable reconfigurability while addressing practical fabrication challenges . To validate the proposed concepts, the designed unit cells are extended into array structures, where their performance is evaluated through both simulation and experimental testing. A practical prototype of a 1-bit reflectarray is fabricated and tested in an anechoic antenna chamber. The prototype demonstrates the key required functionalities, including beam steering, wideband operation, and dual-polarization control. These results confirm the potential of RIS to dynamically manipulate electromagnetic propagation in real-world scenarios. Furthermore, the thesis addresses critical implementation issues related to the scalability of RIS, the integration of control circuitry, and the trade-offs between design complexity and achievable performance. The findings presented in this research underscore the innovative role of RIS in reshaping the architecture of wireless communication systems. By turning the propagation environment into an intelligent and programmable medium, RIS has the potential to significantly improve spectral efficiency, energy utilization, and overall network adaptability. The contributions of this thesis extend the understanding of RIS operation, provide novel unit cell structures, and deliver practical insights for prototyping and implementation. In doing so, this work not only advances academic knowledge in the field but also offers practical guidelines for industrial adoption of RIS in future wireless systems. Ultimately, this research highlights the promise of RIS as a cornerstone technology for realizing the vision of 6G and beyond.
A Study on the Impacts of Wide-Bandgap Devices on Turn-to-Turn Insulation Performance in Hairpin Winding for Electric Vehicle Traction Motors
(University of Waterloo, 2025-12-03) Kholgh Khiz, Ali
Electric Vehicles (EVs) are one of the important levers of transportation electrification. However, charging time and limited mileage per charge remain significant barriers to widespread EV adoption. Continuous advancements in EV technology aim to mitigate these challenges. Three major improvements addressing these issues include increasing the operating voltage level, replacing random-wound motors with hairpin winding, and utilizing wide-bandgap (WBG)-based power converters to drive the motors. These advancements, however, raise concerns regarding motor reliability, particularly in the winding insulation system. Therefore, it is crucial to study and characterize the effects of higher voltage levels and WBG device-based drives on hairpin winding insulation. Turn-to-turn insulation is the most vulnerable point in motor stators. Power electronic converters employ pulse-width modulation (PWM) techniques to generate AC output waveforms, producing pulses with fast (short) rise times, overshoot, high frequency, and variable duty cycles. These PWM-driven systems subject insulation to greater electrical stress than conventional AC-fed machines. The adoption of WBG device-based drives exacerbates this stress due to their inherently fast switching characteristics and high-frequency components. Increased electrical stress may lead to partial discharge (PD) activity, which accelerates insulation degradation. Consequently, evaluating turn-to-turn insulation under WBG device-based drive operation and PD exposure is critical. This study develops a high-voltage SiC-MOSFET pulse generator to investigate the impact of WBG device-based drives on turn-to-turn insulation. A comprehensive analysis is conducted by examining the effects of pulse rise time, overshoot, frequency, and duty cycle. Three rise times (40 ns, 500 ns, and 800 ns) are considered to assess the influence of fast-switching transients inherent to WBG devices. Overshoot effects are examined using 10% and 20% overshoot pulses, while frequency effects are evaluated at 5 kHz and 10 kHz. Additionally, the impact of duty cycle is studied at 20% and 50%. Since traction motors operate under elevated thermal conditions, this study also evaluates the effect of increased temperature on insulation degradation to more accurately replicate in-service stress conditions. To assess turn-to-turn insulation performance, back-to-back test samples replicating hairpin winding structures are developed using actual flat wires employed in EV motors. Two wire types, corona-resistant and non-corona-resistant, are evaluated and compared. Experimental tests are designed based on Design of Experiment (DOE) principles, with samples subjected to 24-hour aging under high-voltage pulses generated by the SiC-MOSFET pulse generator in the presence of PD activity. Insulation performance is assessed before and after aging by measuring partial discharge inception voltage (PDIV) and conducting dielectric frequency response (DFR) analysis. Wire surface temperature is continuously monitored during the aging process, and PD activity is confirmed through the detection of PD electromagnetic wave emissions using a UHF antenna. Furthermore, optical microscopy, atomic force microscopy (AFM), scanning electron microscopy (SEM) images, and EDX analysis are utilized to examine wire coating integrity before and after aging. Results indicate that corona-resistant wires exhibit superior performance under PD conditions compared to non-corona-resistant wires. Additionally, frequency is identified as the dominant factor influencing PDIV drop, whereas overshoot has the most significant effect on the increase in dissipation factor after aging. Microscopy, AFM, SEM, and EDX analysis reveal clear evidence of PD-induced wire coating damage. The combined impact of thermal and electrical stress is examined, with findings compared to room-temperature test results. This research provides critical insights into the reliability of turn-to-turn insulation in hairpin-wound EV motors under WBG device-based drive operation, offering valuable guidance for motor reliability improvement in next-generation EV powertrains.
Zero-Knowledge Proof-Enabled SAT Co-processor for Blockchain Systems
(University of Waterloo, 2025-11-26) Yusiuk, Vladyslav
This thesis explores the possibility of building classical SAT solvers in Circom Domain Specific Language to create zero-knowledge proofs (ZKPs) usable in blockchain contexts. I implemented DPLL and Chaff as arithmetic circuits within Circom and analyze them based on constraint count, proving delay, and zk-SNARK verification layers. With this evaluation, the aim is to determine the feasibility of solvers integration into off-chain computation systems and rollup-centric architectures on Ethereum. The findings indicate that incorporating SAT solvers within zero-knowledge circuits is achievable though some degradation in efficiency occurs based on algorithm used and input representation. This research provides a thorough assessment of known SAT methods across an unconventional boundary, linking symbolic logic with blockchain technologies reliant on zk-SNARKs.
Design and Implementation of a Robust State of Charge Estimation Approach for a Single Battery Cell, a Hardware-in-the-Loop Test Bench, and a Battery Disconnect Unit for an Electric Vehicle Battery Pack
(University of Waterloo, 2025-11-25) Pham, Nguyen Truong Son
As transportation electrification accelerates, battery-powered vehicles, including cars, airplanes, and boats, are rapidly emerging. This thesis provides solutions and practical insights on two key topics: implementing robust machine learning algorithms on commercial Battery Management System (BMS), and building a high-performance Battery Disconnect Unit (BDU). The experience was gained during participation in the North American Battery Workforce Challenge. First, two machine learning approaches for State of Charge (SoC) estimation are introduced. The first approach is an adaptive algorithm using SoC-OCV-T (State of Charge-Open Circuit Voltage-Temperature) lookup table and Extreme Learning Machine (ELM). The experiment began at 100% SoC, with temperature ranging from -20°C to 60°C. From -20°C to 0°C, the maximum absolute error (MAE) ranged from 0.030 to 0.025. In the mid-range from 5°C to 40°C, the MAE decreased to within 0.015 to 0.020 range. Lastly, at higher temperature range of 45°C to 60°C, the MAE was below 0.013. In the second approach, advanced differential features are added to improve the accuracy of the ELM model, particularly below 0°C. Under noisy condition, both the maximum absolute error (MAE) and the root mean square error (RMSE) were reduced to below 1.5% at -20, 20, and 60°C. Both algorithms were validated on a customized Hardware-in-the-loop (HIL) test bench. The HIL platform was developed to streamline validation of algorithms such as SoC estimation. Finally, the thesis details the design and testing process for the BDU, highlighting key design considerations, test results, and engineering challenges.
Mitigating Hardware Trojan Risks in the Global IC Supply Chain: Pre- and Post-Silicon Detection Approaches
(University of Waterloo, 2025-11-19) Pintur, Michael
The integrity of modern systems is critically dependent on trust in the underlying hardware, yet complex Integrated Circuit (IC) supply chains introduce numerous vulnerabilities for malicious insertions. This thesis confronts the challenge of IC trust by examining two distinct detection methodologies, illuminating the fundamental trade-offs inherent in practical hardware verification under black-box conditions. The first contribution targets Trojan detection in Third Party Intellectual Property (3PIP) by adapting power-based side-channel fuzzing with Field-Programmable Gate Arrays (FPGAs). This investigation confirms that dynamic power analysis serves as an effective oracle for identifying the activation of a Trojan, creating a statistically significant side-channel anomaly. However, the work also demonstrates that random fuzzing is an impractical search strategy for discovering the low-probability trigger required for activation, highlighting a significant barrier to its widespread adoption. To overcome the limitations of methods requiring dynamic Trojan activation, this work explores static, on-chip sensing using Ring Oscillator Networks (RONs). This research addresses a gap in prior work by characterizing RON behaviour on a modern 28nm process and subsequently developing a statistical framework to distinguish malicious modifications from normal process variations. The proposed approach was validated against a benchmark hardware Trojan and successfully classified all Trojan-free and Trojan-infected devices. These results confirm that RON-based detection remains effective on 28nm process technology and demonstrate the robustness of the developed anomaly detection algorithm. By juxtaposing a dynamic, trigger-based detection method with a static, reference-based approach, this thesis illuminates the fundamental trade-offs inherent in hardware trust verification. The findings reveal a practical difference between the high specificity of dynamic analysis and the broad applicability of static verification. This research concludes that while physical side-channels are powerful tools, future progress will depend on developing solutions that effectively balance these competing demands, for a more comprehensive security strategy in the IC supply chain.
Fabrication of High Aspect Ratio Atomic Force Microscope Tips Using Silicon and Diamond
(University of Waterloo, 2025-11-18) Nasir, Farheen
Atomic Force Microscope (AFM) is in a family of microscopy called Scanning Probe Microscope (SPM) which also includes Scanning Tunneling Microscope (STM). AFM has become a standard characterization tool for surface topography measurements in most microfabrication. However, in order to obtain accurate representation of high aspect ratio structures such as trenches, the AFM tip must be capable of reaching the bottom of such a trench. This requires the AFM tip to have a high aspect ratio as well as sharp apex. Additionally, for application which require contact with the surface the AFM tip must have very high wear resistance, which is typically obtained using diamond AFM tips . In this work we have designed several methods to obtain high aspect ratio diamond and silicon AFM tips with sharp apex. Chapter 3 and 4 discuss two methods to obtain high aspect ratio diamond tip with silicon base. We have used ultra-nanocrystalline diamond grown on silicon wafer and have fabricated diamond AFM tips using Reactive Ion Etching. In the first method we use O2/CHF3 gas to etch patterned diamond with shrinking mask. Next, we use pseudo-Bosch etching to fabricate high aspect ratio silicon pillars under the diamond tip. With this method we are able to obtain aspect ratio close to 10 and tip apex of less than 25 nm. In the second method we are able to obtain even smaller apex by fabricating the silicon base first before the fabrication of diamond apex using O2/C4F8 plasma. This method allows for aspect ratio of 7.5 and apex diameter close to 10 nm. Chapter 5 and 6 focus on the fabrication of silicon AFM tips. In chapter 5 we use (111) oriented silicon wafers, which are first dry etched to form pillars. Next, the sample is etched in alkaline solution using either TMAH or surfactant added TMAH. We have tested two non-ionic surfactants and have obtained aspect ratio close to 14 and tip apex of less than 10 nm without oxidation sharpening when 1000 ppm Triton X-405 is added to 25 wt% TMAH. We have also studied the effect of surfactant composition and concentration on silicon etching. In chapter 6, we have developed two method to fabricate tetrahedral AFM tips using a combination of dry and wet anisotropic etching. The first method uses SOI wafer coated with LPCVD nitride. After patterning the silicon is dry etched to form trenches which are then oxidized. Next, the LPCVD nitride is removed to exposed un-oxidized silicon that is anisotropically etched in 25 wt% TMAH to form AFM tips with one side perpendicular to the edge of the cantilever surface. In the second method we remove the need of LPCVD nitride which may not be readily available. In this method, after fabrication of trenches, the entire silicon surface is oxidized along with trench sidewalls. However, before anisotropic wet etching, the surface oxide is selectively removed using angled ion milling. Careful recipe optimization is done in order to fully remove the surface oxide while the oxide on the trench sidewalls remain unharmed. After wet anisotropic etching, this method allows for fabrication of silicon tips with apex as low as 60 nm before oxidation sharpening.
Study of Deep Learning Architectures for Bearing Fault Diagnosis Using STFT Spectrograms
(University of Waterloo, 2025-11-18) Simhadri, Rajshri
This thesis presents a comprehensive study on vibration-based bearing fault type and severity-level detection, this process plays a critical role in predictive maintenance for industrial systems. The proposed framework combines signal processing and image-based representations derived from short-time Fourier transform (STFT) spectrograms to classify ten fault classes encompassing various fault types and severities. Among the evaluated architectures, the pretrained ImageNet model XceptionNet-71, when fine-tuned on grayscale STFTs, achieved the best overall performance, attaining a macro F1-score of 0.9979 and a mean ROC–AUC of 0.99 across all classes. This single-channel model demonstrated superior class separability compared to both flattened 1D STFT inputs and three-channel spectrograms. Which confirms that spectrogram-based representations combined with pretrained convolutional backbones are well-suited for bearing fault diagnosis and real-time deployment. While prior studies on the CWRU dataset have improved bearing fault classification through handcrafted features, lightweight CNNs, and transformer-based models, they often suffer from dataset leakage and lack systematic benchmarking. This work addresses these gaps through a unified and reproducible framework that compares 1D and 2D CNNs, extends Delta-STFT into a cross-resolution multi-channel representation, and conducts a comprehensive evaluation to classify safe versus unsafe misclassifications, bridging the gap between high accuracy and practical deployability.
Systematic Methodology for Enhanced Performance Prediction in Designing Large Active Phased Array Antennas for Satellite Communication
(University of Waterloo, 2025-11-14) Tung, Justin
The deployment of 5th generation (5G) and the development of 6th generation (6G) networks have increased the demand for global connectivity at high data rates. Traditional terrestrial infrastructure, including base stations and optical fiber, faces challenges in remote or sparsely populated regions, such as mountainous terrain and oceans. Satellite communication (SATCOM) provides a complementary solution, enabling direct wireless links between users and satellite constellations, removing the need for implementing optic fiber cables. Low earth orbit (LEO) satellites have been proposed as they reduce latency and power requirements compared to traditional geostationary satellites, but their constant motion necessitates rapid beam tracking. At the same time, Ka-band operation (27.5–31 GHz) offers wide bandwidths for high data rates but requires high-gain antennas to overcome propagation losses. Electronically steerable Phased Array Antennas (PAAs) have emerged as the leading solution to address these requirements by enabling fast beam steering without mechanical components. Achieving the necessary gain, however, demands large arrays, which introduces significant design challenges. This thesis presents a systematic design methodology for large 26.5–40 GHz (Ka)-band active PAAs and their feed networks. The methodology leverages industry-standard eletromagnetic (EM) simulation models to guide the design of antenna elements and feed networks, ensuring wideband and scalability. Using this approach, a 16 × 16 active PAA is designed, integrating 64 beamforming integrated circuits (BFICs). Critical system level considerations, including DC power delivery, digital signal integrity, and thermal management, are then analyzed to ensure reliable operation.
Polynomial Controllers for Optimal Trajectory Matching with Stability Guarantees
(University of Waterloo, 2025-11-04) Kitaev, Alexander
We formulate a trajectory matching problem in which a set of reference trajectories for a plant is given, and a control law that causes the plant’s trajectories to be as close as possible to the reference trajectories is desired. These trajectories might be generated by an implicit controller such as a model predictive control (MPC) algorithm or manually chosen by a user. This thesis presents a nonconvex optimization approach for solving the trajectory matching problem that generates explicit polynomial controllers. The value of this approach is that the explicit control laws it generates are simpler to implement, and can be used for stability analysis. Additionally, the method presented in this thesis guarantees local stability of the generated controller by ensuring local contractivity towards the generated trajectories. This thesis presents several theoretical results that justify the method described here. Firstly, a proof that the local contractivity constraints used to ensure local stability can be expressed as a set of matrix inequalities is presented, which turns an infinite set of constraints into a finite one. Secondly, a theorem that describes how symmetries in the trajectory matching problem correspond to symmetries in its solution is presented and proven, which enables a reduction in the control design problem size and resulting solution. Finally, this thesis demonstrates the method it describes on two example problems motivated by real-world applications. The first of these is stabilization and disturbance recovery for a single-machine infinite-bus (SMIB) power system, and the second is a lane change manoeuvre for Dubin’s vehicle, a simple vehicle model. In each case, the reference trajectories are generated by MPC.
An Empirical Study of Privacy Leakage Vulnerability in Third-Party Android Logs Libraries
(University of Waterloo, 2025-10-21) ZHAO, YIXI
Mobile logging libraries are essential tools for debugging and monitoring Android applications, yet their privacy implications remain largely unexplored. This paper presents the first large-scale empirical study of privacy risks in Android logging practices, analyzing 48,702 applications from Google Play to identify sensitive data leakage through third-party logging frameworks. Our findings reveal that while logging library adoption is limited (3.4% of applications), nearly half (49.3%) of logging-enabled applications exhibit privacy leaks, creating significant security vulnerabilities. Three dominant libraries—Timber (35.2%), SLF4J (35.1%), and Firebase (29.4%)—account for 99.7% of all verified privacy leakage instances. We identify distinct logging patterns across frameworks, with SLF4J showing balanced log level distribution, Timber concentrating heavily on DEBUG levels (78.5%), and Firebase dominated by Analytics Events (98.0%). Our analysis reveals that privacy violations predominantly stem from indirect data flows (62.5%) requiring intermediate processing steps, with most leaks occurring through moderate-complexity paths of 2-4 statements. User-info sources dominate privacy leaks (69.7%), while user-input sources represent a substantial portion (30.3%), highlighting GUI components as significant risk vectors. Longitudinal analysis of application updates demonstrates that privacy leaks tend to improve over time, indicating growing developer awareness of privacy concerns, though persistent vulnerabilities underscore the need for systematic privacy protection measures. Our study contributes the largest dataset of third-party logging-based privacy violations to date, a reproducible analysis pipeline for future research, and actionable insights for developers and library maintainers. These findings emphasize the critical need for practitioners to recognize both user information and user input as significant privacy threats when implementing third-party logging frameworks in Android applications.
Radio Resource Management of Hybrid Beamforming Systems
(University of Waterloo, 2025-10-14) Quan, Yuan
Beamforming and massive multi-input multi-output (MIMO) are two of the key technologies that enable high capacity and spectrum-efficient communications in 5G and beyond systems. Codebook-based HBF (Hybrid Beamforming) wherein ABF (Analog Beamforming) vectors are chosen from pre-designed codebooks and, optionally, DBF (Digital Beamforming) can be performed on the selected ABF vectors, results in lower hardware cost, training overheads, and complexity for real-time operations over FDBF (Full Digital Beamforming). We study RRM (Radio Resource Management) for the DL (Downlink) and the UL (uplink) of codebook-based HBF systems assuming proportional fairness. Our study focuses on the practical multi-channel case without assuming that the number of RF (Radio Frequency) chains at the BS (Base Station), K, is larger than the number of UE (User Equipment) in the cell, U. Indeed, in the highly practical yet underexplored case where K
Controlling Light with Photon Subtraction via the Single-Photon Raman Interaction
(University of Waterloo, 2025-10-14) Pasharavesh, Abdolreza
This dissertation leverages deterministic photon subtraction based on the single-photon Raman interaction (SPRINT) to engineer multiphoton quantum fields and design quantum optical platforms for applications ranging from non-Gaussian quantum light generation to photon-number-resolving (PNR) detection and photon number splitting (PNS) attacks on quantum key distribution (QKD). The work is structured into four main parts. In the first part (Chapter 2), we evaluate the performance of the subtraction scheme using system parameters that are technologically accessible according to the current state of the art. We analyze the photon subtraction process in a configuration where the transitions of a Λ-type emitter are selectively coupled to the stationary modes of a bimodal cavity, which are in turn coupled to distinct waveguide modes. Using the input-output formalism of quantum optics and quantum trajectory methods, we investigate single- and multiphoton transport in the system. The results indicate that success rates approaching unity are achievable with currently reported coupling rates for cold atoms trapped in crossed optical-fiber cavities as well as for solid-state platforms based on quantum dots. In the second part (Chapter 3), we explore the capability of the photon subtraction scheme to generate non-Gaussianity in initially Gaussian input fields. Using a photon subtractor with the emitter directly coupled to a chiral waveguide, we show that for both squeezed vacuum and coherent light input pulses, the Wigner function of the output field clearly reveals its non-Gaussian character following photon subtraction. Furthermore, we propose a measurement-based scheme on the subtracted photon which can lead to conditional generation of quantum states resembling Schrodinger’s kitten state directly from coherent input light with fidelities above 99%. This result is particularly noteworthy, as coherent pulses, unlike the squeezed vacuum inputs commonly used in previous studies, are readily available experimentally. The last two parts of the dissertation explore the possibilities arising from cascading multiple photon subtractors. In the third part (Chapter 4), we investigate the operation of a PNR detector composed of a cascade of waveguide-coupled Λ-type emitters, which deterministically demultiplexes incoming photons among single-photon detectors. We present a closed-form expression for the detector’s precision in the linear regime and predict how correlations generated by nonlinear photon-photon interactions influence this precision. We compare the performance of the proposed PNR detector with that of a conventional PNR scheme based on spatial demultiplexing via beamsplitters. Our results indicate that the proposed scheme can outperform conventional detectors under realistic conditions, making it a promising candidate for next-generation PNR detection. In the fourth part (Chapter 5), we present a specialized photon subtraction scheme that enables the deterministic extraction of single photons from multiphoton states while leaving input single-photon states unaltered. The proposed device consists of a two-way cascade of two Λ-type emitters coupled via a chiral waveguide. We analyze the interaction of this system with traveling few-photon pulses of coherent light and use these results to highlight how this two-emitter extension improves the original deterministic single-photon subtraction when it comes to implementing undetectable PNS attack on a QKD channel. Finally, in Chapter 6, we demonstrate how this two-emitter approach can be extended to an n-emitter cascade, resulting in a photon subtractor that selectively extracts photons from an input light stream based on their arrival time sequence. We show that this photon subtractor enables the generation of high-fidelity and modal purity multiphoton Fock states. The application of these Fock-state pulses in optical interferometry is investigated, highlighting their potential for quantum metrology at the Heisenberg limit. These results introduce novel applications of SPRINT-based photon subtraction in areas ranging from non-Gaussian photonics, to PNR detection, QKD, and quantum metrology.
Approaching Memorization in Large Language Models
(University of Waterloo, 2025-10-08) Cheng, Xiaoyu
Large Language Models (LLMs) risk memorizing and reproducing sensitive or proprietary information from their training data. In this thesis, we investigate the behavior and mitigation of memorization in LLMs by adopting a pipeline that combines membership inference and data extraction attacks, and we evaluate memorization across multiple models. Through systematic experiments, we analyze how memorization varies with model size, architecture, and content category. We observe memorization rates ranging from 42% to 64% across the investigated models, demonstrating that memorization remains a persistent issue, and that the existing memorization-revealing pipeline remains valid on these models. Certain content categories are more prone to memorization, and realistic usage scenarios can still trigger it. Finally, we explore knowledge distillation as a mitigation approach: distilling Llama3-8B reduces the extraction rate by approximately 20%, suggesting a viable mitigation option. This work contributes a novel dataset and a BLEU-based evaluation pipeline, providing practical insights for research on LLM memorization.
Enhancing Social Learning in Humanoid Robots Taught by Non-Expert Human Teachers
(University of Waterloo, 2025-09-30) Aliasghari, Pourya
Tools that assist with daily tasks are valuable. For example, with the aging population in Canada and worldwide, there is a growing demand for ways to help older adults perform daily activities independently. Socially intelligent robots can promote independence by assisting with routine tasks. While advanced robots may be capable of performing various specialized operations, it is not feasible for their designers to program them in advance to effectively carry out multi-step, complex tasks requiring high-level planning and coordination, `out of the box' in new environments and for users with diverse preferences. To successfully integrate into domestic environments, robots must learn new task knowledge from human users. Many of our own skills as human beings have been acquired through social learning, i.e., learning via observation of or interaction with others, throughout our lifetime. Social learning for robots enables the transfer of skills without the need for explicit programming, allowing users to teach robots via natural, intuitive, and interactive methods. This thesis targets three key challenges in the social learning of robots: enabling non-expert humans to teach robots without external help, enabling robots to learn and perform multi-step tasks, and enabling robots to identify the most suitable teachers in their social learning. The first phase of my research examines whether or not participants with no prior experience teaching a robot could become more proficient robot teachers through repeated human-robot teaching interactions. An experiment was conducted with twenty-eight participants who were asked to kinesthetically teach a Pepper robot various cleaning tasks across five repeated sessions. Analysis of the data revealed a diversity in non-experts' human-robot teaching styles in repeated interaction. Most participants significantly improved both the success rate and speed of their kinesthetic demonstrations after multiple rounds of teaching the robot. The second phase introduces a novel, biologically inspired imitation approach enabling robots to understand and perform complex tasks using high-level programs that incorporate sequential regularities between sub-goals that a robot can recognize and achieve. To learn a new task, the system processes demonstrations to identify multiple possible arrangements of sub-goals that achieve the overall task goal. For task execution, the robot determines the optimal sequence of actions by evaluating the available sequences based on user-defined criteria, through mental simulation of the real task. This learning architecture was implemented on an iCub humanoid robot, and its effectiveness was evaluated across multiple scenarios. In the third phase, I propose an attribute for identifying the most suitable teachers for a robot: human teachers’ awareness of and attention to the robot’s limitations and capabilities. I investigate the impact of this attribute on robot learning outcomes in an experiment with seventy-two participants who taught three physical tasks to an iCub humanoid robot. Teachers’ awareness of the robot’s visual limitations and learning capabilities was manipulated by offering the robot’s visual perspective and by placing participants in the robot’s position when labelling actions in demonstrations. Participants who could see the robot’s vision output paid increased attention to ensuring that task objects in their demonstrations were visible to the robot. This emphasis on attention resulted in improved learning outcomes for the robot, as indicated by lower perception error rates and higher learning scores. I also propose a metric for robots to estimate the potential for receiving high-quality demonstrations from particular human teachers. These findings demonstrate the feasibility of non-experts adapting to robot teaching through repeated exposure to human-robot teaching tasks, without formal training or external intervention, and also contribute to understanding factors in human teachers that lead to better learning outcomes for robots. Furthermore, I propose a robot learning approach that accommodates variations in human teaching styles, enabling robots to perform tasks with greater flexibility and efficiency. Together, these contributions advance the development of multifunctional and adaptable robots capable of operating autonomously and safely in human environments to assist individuals in various daily activities.

Browse

Recent Submissions