Search results

Title: Improving self-supervised monocular depth estimation from videos using forward and backward consistency
Creator: Shen, Hui
Date: 2020
Description: Recently, there has been a rapid development in monocular depth estimation based on self-supervised learning. However, these existing self...
Show moreRecently, there has been a rapid development in monocular depth estimation based on self-supervised learning. However, these existing self-supervised learning methods are insufficient for estimating motion objects, occlusions, and large static areas. Uncertainty or vanishing easily occurs during depth inferencing. To address this problem, the model proposed in this thesis further explores the consistency in video and builds a multi-frame model for depth estimation; secondly, by taking advantage of the optical flow, a motion mask is generated, with additional photometric loss applied for those masked regions. Experiments are carried out on the KITTI dataset. The proposed model performs better than the baseline model in quantitative results, and as seen from the depth map, the scale uncertainty and depth incomplete situations are improved in motion objects and occlusions explicitly.
Show less

Title: LOW-DOSE CARDIAC SPECT USING POST-FILTERING, DEEP LEARNING, AND MOTION CORRECTION
Creator: Song, Chao
Date: 2019
Description: Single photon emission computed tomography (SPECT) is an important technique in use today for the detection and evaluation of coronary artery...
Show moreSingle photon emission computed tomography (SPECT) is an important technique in use today for the detection and evaluation of coronary artery diseases. The image quality in cardiac SPECT can be adversely affected by cardiac motion and respiratory motion, both of which can lead to motion blur and non-uniform heart wall. In this thesis, we mainly investigate imaging de-noising algorithms and motion correction methods for improving the image quality in cardiac SPECT on both standard dose and reduced dose.First, we investigate a spatiotemporal post-processing approach based on a non-local means (NLM) filter for suppressing the noise in cardiac-gated SPECT images. Since in recent years low-dose studies have gained increased attention in cardiac SPECT owing to its potential radiation risk, to further improve the image quality on reduced dose, we investigate a novel de-noising method for low-dose cardiac-gated SPECT by using a three dimensional residual convolutional neural network (CNN). Furthermore, to reduce the negative effect of respiratory-binned acquisitions and assess the benefit of this approach in both standard dose and reduced dose using simulated acquisitions. Inspired by the success in respiratory correction, we investigate the potential benefit of cardiac motion correction for improving the detectability of perfusion defects. Finally, to combine the benefit of above two types of motion correction, dual-gated data acquisitions are implemented, wherein the acquired list-mode data are further binned into a number of intervals within cardiac and respiratory cycle according to the electrocardiography (ECG) signal and amplitude of the respiratory motion.
Show less

Title: Large-Signal Transient Stability and Control of Inverter-Based Resources
Creator: Wang, Duo
Date: 2024
Description: Renewable generation, including solar photovoltaic (PV) systems, type 3 and 4 wind turbine generation systems (WTG), battery energy storage...
Show moreRenewable generation, including solar photovoltaic (PV) systems, type 3 and 4 wind turbine generation systems (WTG), battery energy storage systems (BESS), as well as high voltage direct current (HVDC) and flexible alternating current (FACT) transmission system devices with increasing penetration level are being connected to the bulk power systems (BPS) via power electronic (PE) converters as the interface, referred to as the inverter-based resources (IBRs) on the transmission and sub-transmission levels or distributed energy resources (DERs) located on the distribution level. The IBR is almost entirely defined by the control algorithms and found to be more prone to experiencing large disturbances due to the lack of the conventional synchronous machine (SM) intrinsic synchronous characteristics and mechanical inertia, as well as being in smaller capacity sizes. Thus, these reasons motivate this dissertation to study the large-signal transient stability and control of IBRs for reliable grid integration and rapid grid transformation. For large-signal stability analysis methods, Lyapunov-based methods are the fundamental theory used to characterize the stability issues with analytical solutions, although other non-Lyapunov methods could also be very helpful. A main difficulty hindering the widespread adoption of the Lyapunov stability analysis method is the difficulty of finding the proper Lyapunov function candidate for a higher dimensional nonlinear system. The Port-Hamiltonian (PH) nonlinear control theory is explored in this dissertation as a promising theoretical framework solution addressing this challenging issue. A PH-based tracking and robust control method is proposed to facilitate the practical application of the PH framework in IBR controls. In addition, considering the typical grid-forming (GFM) IBR control with a first-order low pass filter (LPF) block is usually involved with control saturation function for protection purposes under abnormal operating conditions with anti-windup issue in practical implementation, a PH-based bounded LPF (PH-BLPF) control is proposed to incorporate this in the large-signal PH interconnection modeling framework while preserving the robust tracking Lyapunov stability with improved transient dynamic performance and stability margin.Moreover, specific real-world transient synchronization stability issues, such as the grid voltage large fault disturbance case, are studied. In addition, to meet the recent emerging IBR grid code requirements, such as the current magnitude limitation, grid support function, and fault recovery capability of GFM-VSCs, a virtual impedance-based current-limiting GFM control with enhanced transient stability and grid support is proposed.
Show less

Title: Empowering Visually Impaired Individuals With Holistic Assistance Using Real-Time Spatial Awareness System
Creator: Yu, Xinrui
Date: 2024
Description: The integration of artificial intelligence (AI) into daily life opens unprecedented avenues for enhancing the experiences of visually impaired...
Show moreThe integration of artificial intelligence (AI) into daily life opens unprecedented avenues for enhancing the experiences of visually impaired individuals, offering them greater autonomy and quality of life. This thesis introduces a Visually Impaired Spatial Awareness (VISA) system designed to assist visually impaired individuals holistically through a structured approach. At the foundational level, the VISA system incorporates several key technologies to interpret the surroundings and assist in basic navigation tasks. It utilizes Augmented Reality (AR) markers to facilitate recognition of places and aid in navigation, employs neural network models for advanced object detection and tracking, and leverages depth information for accurate object localization. Progressing to the intermediate level, the VISA system integrates the data obtained from object detection and depth sensing to assist in more complex navigational tasks such as obstacle avoidance and pathfinding toward a desired destination. At the advanced level, the VISA system synthesizes the capabilities developed at the foundational and intermediate levels to enhance the spatial awareness of visually impaired users, allowing them to undertake complex tasks, such as navigating complex environments and locating specific items. The VISA system also emphasizes efficient human-machine interaction, incorporating text-to-speech and speech-to-text technologies to facilitate natural and intuitive communication between the user and the system. The VISA system's performance was evaluated in different environments simulating real-world scenarios. The experimental results show that the user can interact with our system intuitively with minimal effort, and affirm that the VISA system can effectively assist the visually impaired user in locating and reaching for objects, navigating indoors, identifying merchandise, and recognizing both handwritten and printed texts.
Show less

Title: Adaptive Learning Approach of a Domain-Aware CNN-Based Model Observer
Creator: Bogdanovic, Nebojsa
Date: 2023
Description: Application of convolutional neural networks (CNNs) for performing defect detection tasks and their use as model observers (MO) has become...
Show moreApplication of convolutional neural networks (CNNs) for performing defect detection tasks and their use as model observers (MO) has become increasingly popular in the medical imaging field. Building upon this use of CNN MOs, we have trained the CNNs to discern between the data it was trained on, and the previously unseen images. We termed this ability domain awareness. To achieve domain awareness, we are simultaneously training a new variation of U-Net CNN to perform defect detection task, as well as to reconstruct a noisy input image. We have shown that the values of the reconstruction mean squared error can be used as a good indicator of how well the algorithm performs in the defect localization task, making a big step towards developing a domain aware CNN MO. Additionally, we have proposed an adaptive learning approach for training these algorithms, and compared them to the non-adaptive learning approach. The main results that we achieved were for the ideal observers, but we also extended these results to human observer data. We have compared different architectures of CNNs with different numbers and sizes of layers, as well as introduced data augmentation to further improve upon our results. Finally, our results show that the proposed adaptive learning approach with introduced data augmentation drastically improves upon the results of a non-adaptive approach in both human and ideal observer cases.
Show less

Title: Voxel Transformer with Density-Aware Deformable Attention for 3D Object Detection
Creator: Kim, Taeho
Date: 2023
Description: The Voxel Transformer (VoTr) is a prominent model in the field of 3D object detection, employing a transformer-based architecture to...
Show moreThe Voxel Transformer (VoTr) is a prominent model in the field of 3D object detection, employing a transformer-based architecture to comprehend long-range voxel relationships through self-attention. However, despite its expanded receptive field, VoTr’s flexibility is constrained by its predefined receptive field. In this paper, we present a Voxel Transformer with Density-Aware Deformable Attention (VoTr-DADA), a novel approach to 3D object detection. VoTr-DADA leverages density-guided deformable attention for a more adaptable receptive field. It efficiently identifies key areas in the input using density features, combining the strengths of both VoTr and Deformable Attention. We introduce the Density-Aware Deformable Attention (DADA) module, which is specifically designed to focus on these crucial areas while adaptively extracting more informative features. Experimental results on the KITTI dataset and the Waymo Open dataset show that our proposed method outperforms the baseline VoTr model in 3D object detection while maintaining a fast inference speed.
Show less

Title: Application of Blockchain and Artificial Intelligence Methods in Power System Operation and Control
Creator: Farhoumandi, Matin
Date: 2023
Description: The proliferation of distributed energy resources (DERs) and the large-scale electrification of transportation infrastructure are driving...
Show moreThe proliferation of distributed energy resources (DERs) and the large-scale electrification of transportation infrastructure are driving forces behind the ongoing evolution for transforming traditionally passive consumers into prosumers (both consumers and producers) in a coordinated system of power distribution network (PDN) and urban transportation network (UTN). In this new paradigm, peer-to-peer (P2P) energy trading is a promising energy management strategy for dynamically balancing the supply and demand in electricity markets. In this thesis, we propose the applications of artificial intelligence technology to power system operation and control. First, blockchain (BC) is applied to electric vehicle charging station (EVCS) operations to optimally transact energy in a hierarchical P2P framework. In the proposed framework, a decentralized privacy-preserving clearing mechanism is implemented in the transactive energy market (TEM) in which BC’s smart contracts are applied in a coordinated PDN and UTN operation. The effectiveness of the proposed TEM and its solution approach are validated via numerical simulations which are performed on a modified IEEE 123-bus PDN and a modified Sioux Falls UTN. Second, machine learning and deep learning methods are applied to short-term forecasting of non-conforming net load (STFNL). STFNL plays a vital role in enhancing the secure and efficient operation and control of power systems. However, power system consumption is affected by a variety of external factors and thus includes high levels of variations. These variations cause STFNL to be a challenging task as more DERs are integrated into the power grid. This thesis proposes two commonly used machine learning and deep learning methods, i.e., ensemble bagged and long short-term memory, for STFNL. The advantages, features and applications of these methods are expanded in a proposed fusion forecasting model that improves the STFNL accuracy. Additionally, data engineering and preprocessing options are used to increase the accuracy of the proposed fusion model. A comparative study based on practical load data is performed to demonstrate that the proposed fusion methodology can reach a relatively higher forecasting accuracy with lower error indices. Index Terms—Blockchain, deep learning and machine learning, electric vehicle charging stations, non-conforming net load forecasting, peer-to-peer transactive energy, power distribution and transportation networks, distributed energy resources, behind-the-meter supply resources.
Show less

Title: Transactive Energy Market for Electric Vehicle Charging Stations in Constrained Power and Transportation Networks
Creator: Affolabi, Larissa Arielle Sèfiath
Date: 2023
Description: In response to the urgent need for decarbonization, our society is actively working towards reducing carbon emissions across various sectors....
Show moreIn response to the urgent need for decarbonization, our society is actively working towards reducing carbon emissions across various sectors. These efforts have resulted in the widespread adoption of distributed energy resources (DERs) in the electricity sector and the widespread adoption of electric vehicles (EVs) in the transportation sector. The growing popularity of EVs has resulted in rapid growth of charging infrastructure to meet the increasing demand. Recently, combined efforts across those two sectors have gained popularity with the deployment of EV charging stations (EVCSs) with on-site DERs like solar photovoltaic and/or battery energy storage systems not only to defer or avoid the need for power distribution equipment upgrades but also to achieve more environmentally friendly outcomes in terms of decarbonization goals. To increase transportation electrification, we need to expand further the charging infrastructure. The key challenge lies in accelerating charging station deployment while ensuring the safe and efficient operation of the power distribution system where most of this new load will be concentrated. Numerous research efforts have been dedicated to the study of EVCSs, with a focus on either optimizing the pricing of charging services or addressing the energy management challenges from the perspective of system operators. While these aspects are crucial, it is essential to recognize the importance of attracting private sector stakeholders to invest in and support the expansion of the EVCS network. Relying solely on subsidies is insufficient to finance the necessary scale of EVCS deployment required to accelerate the widespread adoption of EVs. The increasing adoption of EVCSs integrated with on-site DERs highlights the potential for Transactive Energy Market (TEM) operations among EVCSs. However, unlike regular prosumers, EVCS operations are uniquely influenced by both the power distribution and the transportation networks. In light of this issue, this dissertation proposes several multi-agent frameworks that leverage on-site DERs at EVCSs to establish a secondary revenue stream through a TEM. This dissertation investigates the technical and economic aspects of these multi-agent frameworks. At its core, we propose two holistic frameworks to solve the energy management problem of EVCSs within a TEM environment. Modeled as independent profit-driven entities, each EVCS optimally schedules its operation based on the day-ahead traffic assignment problem solved by the traffic operator agent. For the TEM clearing process, we propose two distinct lines of approach. First, a centralized approach where a single entity assumes both the market operator and grid operator functions. This integrated approach streamlines the decision-making process and ensures coordinated operations between the market and the power grid. Second, a decentralized approach, where separate entities take on the roles of the market operator and grid operator, respectively. This decentralized structure allows for more flexibility and distributed decision-making within the TEM. Furthermore, in contrast to many TEM related studies that overlook the complexity of the power distribution system, we introduce a comprehensive three-phase unbalanced optimal power flow model. This model incorporates features such as network reconfiguration and tap changers, allowing for a more accurate representation and understanding of the power distribution system's operation. Various case studies are used to prove the effectiveness of our proposed lines of approach to EVCSs’ day-ahead energy management problem.
Show less

Title: Machine Learning (ML) for Extreme Weather Power Outage Forecasting in Power Distribution Networks
Creator: Bahrami, Anahita
Date: 2023
Description: The Midwest region experiences a diverse range of severe weather conditions throughout the year. During the warmer months, thunderstorms,...
Show moreThe Midwest region experiences a diverse range of severe weather conditions throughout the year. During the warmer months, thunderstorms, heavy rain, lightning, tornadoes, and high winds pose a threat, while the colder season brings ice storms, snowstorms, high winds, and sleet storms, all of which can cause significant damage to the environment, properties, transportation systems, and power grids. The average climate in the Midwest is influenced by factors such as latitude, solar input, water systems' typical positions and movements, topography, the Great Lakes, and human activities. The combination of these conditions during different seasons contributes to the development of various types of storms. Therefore, it is crucial to predict the impacts of such atmospheric events on distribution and transmission lines, enabling utilities to assess and implement preventive measures and strategies to minimize the economic losses associated with these disasters. Additionally, the accurate classification of storm modes through an automated system allows operators to study trends in relation to climate change and implement necessary strategies to ensure grid reliability and resilience.In recent years, a significant number of power outages have occurred due to extreme ice formation on transmission and distribution networks, posing a threat to the power grid's resilience and reliability. To prepare power providers for snowstorms, extensive research has been conducted on snow accretion on power lines. Over the past two decades, many scientists have turned to machine learning (ML) algorithms for predicting ice accretion on overhead conductors, as ML models demonstrate superior accuracy compared to statistical forecasting models when it comes to forecasting challenging and fine-grained problems. However, most existing models primarily focus on predicting ice formation on power lines and fail to forecast the resulting damage to the distribution network. Therefore, this project proposes a model for predicting power outages caused by snow and ice storms in the distribution network. The goal is to aid in the planning process for disaster response and ensure the resilience and reliability of the power grid. The proposed outage prediction model incorporates statistical and machine learning techniques, taking into account features related to weather conditions, storm events, and information about the power network feeders.
Show less

Title: Optimization of Large-Scale NOMA With Incidence Matrix Design and Physical Layer Security
Creator: Hwang, Eli W.
Date: 2024
Description: The Non-Orthogonal Multiple Access (NOMA) system is recognized for its capability to achieve higher spectral efficiency and massive...
Show moreThe Non-Orthogonal Multiple Access (NOMA) system is recognized for its capability to achieve higher spectral efficiency and massive connectivity. NOMA is intended to transmit massive user communications. The incidence matrix governs the relationship between users and resources for the Code domain NOMA (CD-NOMA). However, NOMA studies focus less on the design and optimization of the incidence matrix.Therefore, this thesis aims to investigate the development of a secure and large-scale NOMA system based on incidence matrix design. The main contributions are outlined as follows: Firstly, this research introduces a novel NOMA system. Distinct from existing studies, the NOMA system is based on combinatorial design. This innovative approach, coupled with a unique constellation design, eliminates the surjective mapping from the linear adding data of multiusers, reducing the complexity of constellation design and Multiuser Detection (MUD). The characteristics of the incidence matrix designs, Simple Orthogonal Multi-Arrays (SOMA), are explored, which display a distinct Latin Square pattern. The SOMA design's unique structure allows for the creation of a highly flexible and fair resource allocation matrix. The NOMA system's theoretical performance analysis equations are established, supporting dynamic adaptability and optimization. The design is validated by Monte Carlo simulation. Compared to other NOMA schemes, it offers higher degrees of freedom and lower complexity while maintaining graceful error rates to transmit a larger number of users. Secondly, a novel NOMA system utilizing incidence matrix information in the uplink is investigated. The incidence matrix pattern is exploited for MUD to achieve large-scale user connectivity. The incidence matrix is designed based on two critical mathematical concepts: parallel classes in hypergraph theory and orthogonal arrays (OAs) in combinatorial designs. Unlike other NOMA schemes, which require modification of their receiver and transmitter to decode superimposed multiuser signals, the unique pattern of the OA structure enables the use of conventional modulators. Consequently, the system load increases and the complexity and latency are reduced. The order of magnitude of the decoding complexity can be significantly reduced from O(N^3) to O(N) compared to the conventional minimum mean-square estimation (MMSE) decoder. Monte Carlo simulation validates that this novel NOMA system outperforms other NOMA designs in terms of error rate, data rate, and system size. Finally, a reconfigurable convolutional encoder design that integrates security and error correction based on physical layer security (PLS) and randomness is developed. This design addresses concerns over privacy, security, and reliability of Internet of Things devices in edge computing networks. The lightweight Convolutional encoders are designed to ensure security by updating the transfer function dynamically with user data. The reconfigurability of the design is achieved by replacing the fixed adder that represents the generator polynomials with the switch adder, enabling the use of 87 billion distinct updating structures, thereby enhancing the versatility of the design. BER-based PLS paradigms are demonstrated in the simulation. In the simulation, the robustness and randomness of this design are further validated through tests suggested by the National Institute of Standards and Technology for cryptographically secure pseudorandom number generators, such as the monobits, longest one, and run tests.
Show less

Title: A Kernel-Free Boundary Integral Method for Two-Dimensional Magnetostatics Analysis
Creator: Jin, Zichao
Date: 2023
Description: Performing magnetostatic analysis accurately and efficiently is crucial for the multi-objective optimization of electromagnetic device designs...
Show morePerforming magnetostatic analysis accurately and efficiently is crucial for the multi-objective optimization of electromagnetic device designs. Therefore, an accurate and computationally efficient method is essential. Kernel Free Boundary Integral Method is a numerical method that can accurately and efficiently solve partial differential equations. Unlike traditional boundary integral or boundary element methods, KFBIM does not require an analytical form of Green’s function for evaluating integrals via numerical quadrature. Instead, KFBIM computes integrals by solving an equivalent interface problem on a Cartesian mesh. Compared with traditional finite difference methods for solving the governing PDEs directly, KFBIM produces a well-conditioned linear system. Therefore, the numerical solution of KFBIM is not sensitive to computer round-off errors, and the KFBIM requires only a fixed number of iterations when an iterative method (e.g., GMRES) is applied to solve the linear system.In this research, the KFBIM is introduced for solving magnetic computations in a toroidal core geometry in 2D. This study is very relevant in designing and optimizing toroidal inductors or transformers used in electrical systems, where lighter weight, higher inductance, higher efficiency, and lower leakage flux are required. The results are then compared with a commercial finite element solver (ANSYS), which shows excellent agreement. It should be noted that, compared with FEM, the KFBIM does not require a body-fitted mesh and can achieve high accuracy with a coarse mesh. In particular, the magnetic potential and tangential field intensity calculations on the boundaries are more stable and exhibit almost no oscillations.Furthermore, although KFBIM is accurate and computationally efficient, sharp corners can be a significant problem for KFBIM. Therefore, an inverse discrete Fourier transform (DFT) based geometry reconstruction is explored to overcome this challenge for smoothening sharp corners. A toroidal core with an airgap (C-core) is modeled to show the effectiveness of the proposed approach in addressing the sharp corner problem. A numerical example demonstrates that the method works for the variable coefficient PDE. In addition, magnetostatic analysis for homogeneous and nonhomogeneous material is presented for the reconstructed geometry, and results carried out from KFBIM are compared with the results of FEM analysis for the original geometry to show the differences and the potential of the proposed method.
Show less

repository.iit

Search the repository

Pages

Pages

Enabled Filters

Refine Results

Type

Date

Department

Subject

Creator

Rights