Department of Electrical and Computer Engineering Graduate Theses
Permanent URI for this collection
Browse
Recent Submissions
Item From Dead-Reckoning to Dynamic State Estimation: Advancements in Autonomous Vehicle Positioning(2024-09-13) Marques de Araujo, Paulo Ricardo; Electrical and Computer Engineering; Givigi, Sidney; Noureldin, AboelmagdThis thesis presents advancements in the field of Autonomous Vehicles (AVs) navigation, focusing on the integration of radar systems, Inertial Measurement Units (IMUs), and advanced learning techniques to enhance state estimation and positioning. The research addresses the limitations of traditional IMU-based methods, which often suffer from noise and bias, leading to rapid error propagation. By introducing the Radar Inertial Dead-Reckoning (RIDR) system, this work effectively reduces these issues, providing more reliable state estimates through the integration of radar-based forward speed measurements and deep learning techniques. The thesis further explores intelligent methods for forward speed estimation using various radar configurations. The development of Deep Neural Network (DNN) models for 1D, 2D, and 3D radar data showcases substantial improvements in noise reduction and estimation accuracy, outperforming traditional approaches in dynamic environments. Additionally, the application of Invariant Extended Kalman Filters (IEKFs) for robust vehicle positioning in urban areas is investigated, demonstrating significant enhancements in accuracy by integrating RIDR with GNSS and map-based corrections. A novel self-supervised network, QuasittudeNet, is introduced for precise attitude estimation using IMU data and forward speed estimates. This network leverages a quasi-static gravity model to accurately determine pitch and roll angles, addressing sensor noise and biases. Furthermore, the integration of Continuous Action Learning Automata (CALA) with IEKFs for dynamic optimization of the measurement covariance matrix is presented. This method significantly improves performance during GNSS outages, maintaining high accuracy and reducing the need for manual tuning. Overall, this research highlights the potential of advanced sensor fusion, deep learning, and adaptive optimization techniques in improving autonomous vehicle navigation systems.Item Design and Control of Parallel Three-Phase AC Motor Drives in Battery-Electric Heavy-Haul Freight Locomotives(2024-09-12) Matthews, Peter Darrach; Electrical and Computer Engineering; Eren, SuzanThe transportation sector is one of the largest contributors to global greenhouse gas emissions. Due to climate change, many companies operating within the transportation sector are seeking to reduce their emissions, including Ontario Northland Rail (ONR), who operate afreight rail network in northern Ontario. ONR currently relies on diesel powered locomotives for their freight rail operations. To reduce ONR’s greenhouse gas emissions, researchers at the ePOWER facility of Queen’s University have proposed the development of a prototype battery-electric locomotive which could work in tandem with existing diesel locomotives to hybridize a freight train. One important component of any electric vehicle is the motor drive, which is the electronic circuit which manages the flow of power between the electric traction motors and the vehicle’s battery. In this thesis, a parallel motor drive system suitable for high power AC traction applications, such as battery-electric locomotives, is proposed. A novel control scheme is developed which is implemented directly in the natural, abc, reference frame and overcomes the primary challenge when designing parallel motor drive systems: circulating currents. At the core of this control strategy is a proposed Resonant Proportional Integral (RPI) controller, which uses integrated plant dynamics to achieve the functionality of a second-order Proportional Resonant (PR) controller using only a first-order Proportional Integral (PI) controller. Hence the proposed control strategy is very simple, requiring only an inner first-order RPI controller for the stator currents and an outer PI controller for motor speed and maximum torque per ampere (MTPA) operation. A theoretical analysis of the controller is given, which shows the control is robust and stable for all expected motor speeds. The proposed motor drive system is then simulated using the proposed controller and a conventional controller. The proposed control system is found to match the dynamic speed and torque performance of the conventional controller while also effectively suppressing the circulating currents. Prototype inverter modules are then designed and used to experimentally validate the proposed control scheme. The experimental results show that the proposed control method achieves the claimed performance.Item MACHINE LEARNING BASED NONLINEAR CHARACTERIZATION IN HETEROGENEOUS OPTICAL NETWORKS(2024-09-11) Boertjes, Matthew; Electrical and Computer Engineering; Cartledge, John; Chan, GeoffreyOptical networks have seen exponential increases in data traffic in recent years due to advanced bandwidth hungry communication services such as 5G, the internet of things, and cloud computing. Such increases cause a great demand for increased capacities within optical networks, where the major factor limiting achievable capacity is fiber nonlinearities. As such, an essential aspect of optical transmission networks is the ability to characterize and subsequently compensate for the effects of fiber nonlinearities. This work investigates the use of analytical, numerical, and machine learning based models in the characterization of the effects of nonlinear distortions. A novel method for building a machine learning based nonlinear signal-to-noise ratio estimator is presented. The proposed model building method aggregates important features from three ensemble models with boosting and shows universal application to all considered training cases. A vast set of heterogeneous system configurations are considered for model training and demonstrate improved model generalization without the need for frequent retraining. A comparative analysis is presented which investigates the agreement of three approximate XPM models in terms of modeling the characteristics of XPM induced phase noise. From this analysis, it is shown that simpler approximate models, while attractive for machine learning dataset generation, struggle when modeling the statistical characteristics of XPM induced phase noise. Investigation into the symbol pattern dependence, manifestation of artifacts in approximate models, and alterations made to each approximate model targeting their limitations is also explored. A method is presented to separate the XPM induced phase noise from a cumulative noise observation which can sufficiently separate a representation of the true XPM induced phase noise. The separation method can be used in machine learning model dataset generation to improve model applicability to practical systems. Results from this work show promise in characterizing nonlinear distortions to assist machine learning based nonlinear compensation models in optical networks.Item Assessment and Mitigation Approaches of 5G C-band Interference with Aeronautical Radar Altimeter(2024-08-28) Elsayem, Aisha; Electrical and Computer Engineering; Noureldin, Aboelmagd; Elghamrawy, HaidyThe recent deployment of 5G technology in the C-band has raised concerns regarding potential interference with aeronautical radar altimeters. 5G systems in the C-band operate within a frequency range of 3.7-3.98 GHz, which closely aligns with the operational frequency of radar altimeters, falling within the range of 4.2-4.4 GHz. This proximity in operational frequencies increases the possibility of interference between the two systems. In this thesis, we explore two primary objectives: firstly, to investigate and test the potential for interference between 5G C-band and radar altimeters, and secondly, to develop methods for interference mitigation. To accomplish this, three interference assessment approaches are explored. The first involves statistical interference analysis with randomly generated base stations (BSs) within the altimeter coverage area. The second develops comparative interference analysis with a single BS radiating close to the touch-down point of aircraft. The final approach assesses interference in a real-life scenario where multiple BSs are implemented to serve a real-world runway. Additionally, two interference management techniques were proposed and evaluated within the assessed real-life scenario. The first involves the implementation of adaptive BS using the power control (PC) method, which aims to mitigate interference with minimal impact on coverage by adjusting the transmitting power for the BS that contributes the most to the interference model. A modification to this technique was applied to loop over the coverage areas instead of individual BSs. This technique is useful in scenarios where BSs are implemented close to each other with overlapping coverage. Lastly, a sequential quadratic programming (SQP) optimization algorithm was developed to optimize the locations of BSs, minimizing interference while maintaining coverage. This work has explored the impact of potential interference between 5G in the C-band and radar altimeters and suggested practical methods to allow the coexistence of both systems, thereby ensuring aviation safety and fulfilling the telecommunication sector’s objectives.Item Prediction of Pneumonia Mortality Risk and Cognitive Test Scores With Interpretable Machine Learning Models(2024-08-28) Sanii, James; Electrical and Computer Engineering; Chan, Wai YipAdopting machine learning algorithms in medical practices is challenging due to their lack of transparency. This thesis shows that interpretable models can offer similar, if not better, performance than traditional machine learning methodologies when applied to tabular data with interpretable features. Confirming that critical variables align with existing medical domain knowledge verifies that the model learns from relevant patterns instead of statistical coincidences or improperly created variables in the data set. Researchers need to investigate high-impact variables that are not known to correlate with a given condition to determine their relevance and improve existing medical domain knowledge. This thesis explores applying explainable machine learning practices to two problems: pneumonia mortality risk prediction and predicting future cognitive test scores. In our first case study, we proposed a novel pneumonia risk prediction framework using an explainable boosting machine model to predict patient mortality risk to optimize hospital resource usage. We pruned the model feature set to only allow for medically relevant features, which offered minimal performance decay while outperforming other machine learning methods. The model outperformed all prior work on the MIMIC-III dataset for this task. Our second case study focused on predicting future cognitive test scores for the Canadian Longitudinal Study on Aging. We pruned the large dataset, which had over 6,000 input variables, down into 25 lightweight, explainable feature models with minimal performance loss from the feature pruning process. Results from this work show that there is promise in using explainable machine learning models to predict future cognitive test scores, which is the first step in applying early preventative measures for irreversible cognitive decline due to dementia or Alzheimer's disease. Both case studies show that explainable machine learning on tabular data offers similar, if not better, results than black models.