University of Illinois Urbana-Champaign
Illinois Digital Environment for Access to Learning and Scholarship RepositoryNot a member yet
123813 research outputs found
Sort by
Urban climate–energy interactions from global to local scales
No full textSubmission published under a 24 month embargo labeled 'Closed Access', the embargo will last until 2027-05-01The student, Xinchang Li, accepted the attached license on 2025-01-09 at 11:33.The student, Xinchang Li, submitted this Dissertation for approval on 2025-01-09 at 11:37.This Dissertation was approved for publication on 2025-01-17 at 13:19.DSpace SAF Submission Ingestion Package generated from Vireo submission #21610 on 2025-10-19 at 19:52:15Earth’s climate and energy systems are closely intertwined through complex interactions. Urban energy use both drives and is affected by anthropogenic climate change. Previous studies have predominantly examined the global scale interactions between energy use and future warmer climates. However, the local scale interactions between urban energy use and the urban climates are frequently ignored in future energy projections, due to methodological, scale, and computational challenges. These local interactions can have global scale effects. Ignoring such interactions underestimates climate-driven energy risks on global to local scales, undermining our climate preparedness in energy planning. This dissertation aims to advance our understanding on the mechanisms and impacts of local urban climate-energy interactions on global-to-local energy demand in a changing climate. I achieve this goal by establishing a hybrid modeling framework integrating process-based modeling and machine learning, and by improving the representation of urban climate-energy interactions in a global Earth system model (ESM). Using the hybrid modeling framework, I first show the prevalent energy projection methods misrepresent the magnitude, nonlinearity, and uncertainty in the climate-driven projections of future building energy demand due to the missing two-way feedbacks between urban climate and energy. I find a 220% increase (47% decrease) in cooling (heating) energy demand with amplified uncertainty by 2099 under a very high emission scenario, roughly twice that projected by previous methods. Cities’ building energy demand responses to future warming climates are spatially diverse, which necessitates urban energy planning accounting for the unique local climate–energy interactions. This work underscores the critical necessity of explicit and dynamic modeling of urban building energy use for climate-sensitive energy planning. Next, I improve the building energy parameterization in a global ESM. I establish a new scheme that represents air-conditioning (AC) adoption explicitly through an AC adoption rate parameter, and build a global, present-day, survey-based, and spatially explicit AC adoption rate dataset at country and sub-country level. The new scheme and dataset significantly improve the accuracy of AC energy demand modeling and enable the evaluation of the effects of changing AC adoption on urban energy and climate across scales through global physics-based dynamic modeling. This work represents continued efforts in better representing urban processes and coupled human-urban-Earth dynamics in ESMs. Finally, I examine the effect of humidity on AC energy demand across global cities under climate change. By modeling building latent heat load as part of AC energy demand in an ESM, I show humidity increases AC energy demand exponentially on hot and humid days and may cause unexpected demand spikes on mildly hot days across diverse climate zones. This effect is further exacerbated by climate change. Divergent humidity-driven shifts will occur in cities’ building energy design space as a result of the interplay of local temperature and humidity changes under climate change. Through development of new model schemes, construction of data products, and hybrid modeling simulations, this dissertation demonstrates the importance of capturing urban climate-energy interactions for comprehensive climate impact assessment, science-based policy-making, and inter-region coordination on climate-sensitive energy planning. This necessitates continued improvement in physical representations of urban energy systems in large-scale models as an important direction of future work
Expanding electronics beyond silicon with wide-bandgap, 2D, and ferroelectric materials
No full textSubmission published under a 24 month embargo labeled 'Closed Access', the embargo will last until 2027-05-01The student, Hanwool Lee, accepted the attached license on 2025-04-17 at 01:00.The student, Hanwool Lee, submitted this Dissertation for approval on 2025-04-17 at 01:08.This Dissertation was approved for publication on 2025-04-18 at 16:08.DSpace SAF Submission Ingestion Package generated from Vireo submission #21812 on 2025-10-19 at 19:53:17This dissertation explores the advancement of microelectronics through novel materials, including two-dimensional (2D) materials, ferroelectric materials, and wide-bandgap semiconductors. These emerging materials enable new functionalities, improve energy efficiency, and enhance stability for various applications. Chapter 1 provides background information for this dissertation, including a brief review of 2D materials, particularly transition metal dichalcogenides (TMDCs). Ferroelectric materials and their device applications are discussed. Additionally, wide-bandgap semiconductors and their advantages are introduced, with a particular focus on gallium nitride (GaN). Chapter 2 explores non-volatile reconfigurable transistors with four-mode operation. Utilizing the strong polarization of epitaxially grown scandium aluminum nitride (ScAlN), a single device can function as an n-type, p-type, always-on, or always-off transistor. The feasibility of these transistors for logic gate applications is demonstrated. Additionally, non-volatile latch operation is presented using van der Waals materials, including ferroelectric copper indium thiophosphate (CIPS) and molybdenum ditelluride (MoTe2). Ferroelectric field-effect transistor (FeFET) with metal-ferroelectric-metal-insulator-semiconductor (MFMIS) structure enables stable memory operation. Using these FeFETs, non-volatile sequential logic operation is demonstrated through a simple latch circuit. Chapter 3 demonstrates the wafer-scale synthesis of MoTe2 using di-tert-butyl telluride ((C4H9)2Te) as the tellurium precursor, along with molybdenum hexacarbonyl (Mo(CO)6) and sputtered molybdenum (Mo) as molybdenum precursors. The successful wafer-scale growth of both 1T' and 2H phases of MoTe2 is presented, with various characterization results confirming the uniformity, phase selectivity, and high crystallinity of the synthesized material. Chapter 4 investigates GaN-based high-electron-mobility transistors (HEMTs) for high-temperature applications. Dielectric stack optimization, gate recess structures, and p-GaN/AlGaN/GaN heterostructures are explored to achieve stable operation up to 500 °C. Optimizing the dielectric stack enhances the breakdown field and device lifetime, while the gate recess and p-GaN/AlGaN/GaN heterostructure enable enhancement-mode operation with improved threshold voltage stability at high temperatures. Chapter 5 concludes this dissertation by summarizing key findings and outlining directions for future research. By integrating emerging materials with innovative design strategies, these studies advance next-generation electronic devices and facilitate their practical implementation in semiconductor technology
Energy-efficient spiking neural network-based visual place recognition and reinforcement learning in adversarial scenarios
No full textSubmission published under a 24 month embargo labeled 'Closed Access', the embargo will last until 2027-05-01The student, Ugur Akcal, accepted the attached license on 2025-04-28 at 08:27.The student, Ugur Akcal, submitted this Dissertation for approval on 2025-04-28 at 08:43.This Dissertation was approved for publication on 2025-04-30 at 07:07.DSpace SAF Submission Ingestion Package generated from Vireo submission #22021 on 2025-10-19 at 19:54:19Spiking neural networks (SNNs) have attracted significant attention as an emerging third-generation artificial intelligence (AI) technology due to their potential for remarkable computational efficiency when deployed on neuromorphic hardware. Research has shown that SNNs can achieve energy efficiencies several orders of magnitude greater than their conventional artificial neural network (ANN) counterparts. Although there are numerous outstanding demonstrations of ANNs, their deployment on platforms with limited computational resources is quite restricted. In this context, SNNs offer considerable potential for expanding the deployment range of AI-based robotics solutions, driving a growing and sustained interest in SNN research. However, training state-of-the-art (SOTA) SNNs tailored for robotics problems is often intractable, and they typically demonstrate poor real-time performance. This motivates the current work to develop SNNs with tractable training that can yield better performance than existing SOTA SNNs in two domains: 1) Visual Place Recognition (VPR) and 2) Reinforcement Learning (RL) in adversarial scenarios. VPR is the ability to recognize locations in a physical environment based solely on visual inputs—a challenging task due to perceptual aliasing, viewpoint and appearance variations, and the complexity of dynamic scenes. To address the shortcomings of existing SNN approaches, we developed an end-to-end convolutional SNN model for VPR that leverages backpropagation for tractable training. During training, we employ rate-based approximations of leaky integrate-and-fire (LIF) neurons, which are then replaced with spiking LIF neurons during inference. The proposed method significantly outperforms existing SOTA SNNs on challenging datasets such as Nordland and Oxford RobotCar, achieving precision at recall on the Nordland dataset (compared to from the current SOTA) and on the Oxford RobotCar dataset (compared to from the current SOTA). Our approach offers a simpler training pipeline while yielding significant improvements in both training and inference times compared to existing SNNs for VPR. Hardware-in-the-loop tests using Intel's neuromorphic USB form factor, Kapoho Bay, show that our on-chip spiking models for VPR trained via the ANN-to-SNN conversion strategy continue to outperform their SNN counterparts, despite a slight but noticeable decrease in performance when transitioning from off-chip to on-chip, while offering significant energy efficiency. The results highlight the outstanding rapid prototyping and real-world deployment capabilities of this approach, showing it to be a substantial step toward more prevalent SNN-based real-world robotics solutions. RL literature has seen a surge in applications of SNNs, due to their computational efficiency when deployed on neuromorphic hardware. Existing work commonly uses population coding, reward-modulated spike timing-dependent plasticity (R-STDP), or other three-factor Hebbian rules. While these methods perform adequately on simple, less stochastic tasks, they falter in more complex and highly stochastic settings, such as a multi-agent Capture-the-Flag (CtF) game. This shortfall arises from various issues, including substantial hyperparameter tuning burdens, vulnerabilities to the dead neuron problem, and vanishing gradients. To address these challenges, we propose S2Act, a computationally lightweight spiking actor-critic network that adopts the ANN-to-SNN conversion paradigm. By employing rate-based approximations of Leaky LIF neurons that mimic Rectified Linear Unit (ReLU) activation functions during training, we mitigate the vanishing gradient problem. After training, we replace the rate-based LIF approximations with the original spiking LIF neurons for inference and deployment on neuromorphic hardware. We evaluated S2Act in a simulated parking task and more challenging multi-agent CtF environment alongside relevant SNN baselines. We also implemented S2Act in a real-world multi-robot CtF demonstration using Intel’s neuromorphic USB form factor Kapoho Bay on TurtleBot platforms. Our experimental results show that S2Act outperforms existing SNN baselines, as it achieves a remarkable improvement in training time and superior task performance with a significantly smaller network size. These findings highlight the potential of S2Act for efficient real-world deployment of RL-based robots in complex tasks
Generative AI and co-creation on social media
No full textSubmission published under a 24 month embargo labeled 'Closed Access', the embargo will last until 2027-05-01The student, Maggie Zhang, accepted the attached license on 2025-05-01 at 11:30.The student, Maggie Zhang, submitted this Dissertation for approval on 2025-05-01 at 11:37.This Dissertation was approved for publication on 2025-05-01 at 16:56.DSpace SAF Submission Ingestion Package generated from Vireo submission #22151 on 2025-10-19 at 19:55:38This dissertation investigates the impact of Generative AI (GAI) on the co-creation of both content and identity on social media. First, this dissertation moves beyond treating humans and AI as distinct entities to propose a blended paradigm of human-AI hybrids, presented as a theoretical framework in Chapter 2. Then, two empirical studies are conducted in Chapter 3 and Chapter 4: the first study investigates how GAI assistance influences the processes and outcomes of content co-creation between humans and AI through field experiments using custom-developed browser extensions on Twitter. The second study explores the delegation of social interactions to personal AI assistants, analyzing observational data from Weibo and examining audience reactions to such identity co-creation. Overall, the findings reveal that GAI can have mixed values for content and identity co-creation
The Catalyst: UIS Research Review, Issue 7
No full textThe Catalyst is a publication by the Research Society at UIS that highlights student research at the university. This issue includes Makenly Jones and her research, Environmental Heavy Metal Analysis Using X-ray Fluorescence. This issue includes a faculty spotlight of Dr. Alex Wolfe and his research on the association between context specific sedentary behaviors and content specific academic achievement
The tactical dialectic: reframing tactical technical communication through the writing practices of activists and technical experts
Get PDFSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-20 without embargo termsThe student, Rebecca Avgoustopoulos, accepted the attached license on 2025-05-07 at 14:31.The student, Rebecca Avgoustopoulos, submitted this Dissertation for approval on 2025-05-07 at 14:40.This Dissertation was approved for publication on 2025-05-12 at 13:19.DSpace SAF Submission Ingestion Package generated from Vireo submission #22213 on 2025-10-20 at 16:57:01This dissertation explores the evolving role of Technical and Professional Communication (TPC) within digital platforms. While traditional TPC focused on institutional writing and technical documentation, the rise of user-generated content and algorithm-driven communication has expanded the field’s boundaries. Tactical Technical Communication (TTC), introduced by Miles Kimball, emphasizes every day, user-created content that resists institutional norms. However, based on Michel de Certeau’s The Practice of Everyday Life., TTC often relies on a binary framework that positions individual tactics in opposition to institutional strategies, This project proposes the Tactical Dialectic (TD) to better capture how technical communicators operate within and alongside institutional systems, especially in online environments. Rather than purely resisting structures, users navigate and adapt them to meet community and personal goals. Case studies include the activist group Emojination, which works within the Unicode Consortium to advocate for more inclusive emojis, and AI/ML researchers who publish in public forums to democratize access to technical knowledge. These examples reveal how technical writers use platform tools strategically while maintaining user agency. The Tactical Dialectic reframes TTC for the digital age, offering a model that accounts for collaboration, adaptation, and resistance within complex, networked systems of communication and control
Networks for motion and motion for networks
Get PDFSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-20 without embargo termsThe student, Emerson Sie, accepted the attached license on 2025-06-12 at 16:35.The student, Emerson Sie, submitted this Dissertation for approval on 2025-06-15 at 12:30.This Dissertation was approved for publication on 2025-06-24 at 11:18.DSpace SAF Submission Ingestion Package generated from Vireo submission #22336 on 2025-10-20 at 16:57:10Wireless systems enable key sensing and connectivity capabilities for mobile platforms. However, existing wireless systems face challenges, preventing important applications from being realized. First, current sensing techniques are too coarse-grained or computationally intensive, limiting practicality for small, agile platforms such as drones and robots. Second, current connectivity systems remain vulnerable to environments where signals are blocked, preventing connectivity in agriculture and geo-location indoors. In this thesis, we address these challenges by exploring the interplay between networks and motion. Specifically, we contribute techniques for improving mobile platforms using networks and improving networks using mobile platforms. First, we contribute a new wireless sensing primitive for agile motion sensing via surface-parallel Doppler shift. We use this to perform accurate odometry and simultaneous localization and mapping on embedded mmWave radars. Next, we expand the scope of rural connectivity for digital agriculture. We contribute an agile and low-cost broadband connectivity model for under-canopy robots operating on remote farms. Our system exploits horizontal and vertical motion of a cellular base station to optimize coverage and throughput to clients. Finally, we tackle geo-location in GPS-denied indoor environments. Although many indoor localization systems have been proposed through the years, they remain too impractical for widespread real-world deployment. To address this, we describe a localization system that works with any unmodified Wi-Fi device. We show how pedestrian crowdsourcing can bootstrap this system in any environment containing enough Wi-Fi APs without dedicated human effort. We believe this is a promising step towards realizing ubiquitous GPS-level localization in urban environments across the world
Life history and the structure and stability of plant communities
Get PDFSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-20 without embargo termsThe student, Kenneth Jops, accepted the attached license on 2025-06-19 at 13:08.The student, Kenneth Jops, submitted this Dissertation for approval on 2025-06-19 at 13:18.This Dissertation was approved for publication on 2025-06-24 at 16:18.DSpace SAF Submission Ingestion Package generated from Vireo submission #22346 on 2025-10-20 at 16:57:12Life history, the pacing of growth, death, dispersal, and reproduction across the life cycle of species, has long been used to classify and differentiate species and to refine predictions on their demographic patterns. This dissertation expands the applications of life history theory to the traditional domain of theoretical ecology, the structure and composition of biological communities. Unlike other mechanisms to explain and predict the stability of communities such as niche partitioning, fitness differences, environmental responses, and direct competitive interaction, a species’ life history can be measured relatively accurately using only a small number of parameters that are often collected in field censuses. Using the familiar language of Matrix Population Models, I will demonstrate that the internal demographic dynamics of populations can influence their impact on larger, diverse communities of multiple species. In the first chapter, I introduce the key parameter for classification of a species’ dynamics in models of life history variation, Ny, the effective population size measured in years. I develop a model of species interaction without direct competition to isolate the effects of life history. I demonstrate how this parameter predicts coexistence timescales for pairwise competition models and governs species richness in models of metacommunities with outside immigration, revealing that life history variation imposes theoretical constraints on species assembly and competition in communities. I also show statistical support for the hypothesis that life history complementarity, a clustering or parity of Ny values in coexisting species, is prevalent across a wide range of sampled populations. In the second, I integrate the mechanisms of niche partitioning and fitness differences into models of variation in life history. I demonstrate how these classical mechanisms can increase persistence times beyond those predicted by life history complementarity alone. I show modeling results for this theoretical framework, and an analytical estimation for competitive ability in limited ranges of fitness variation. This work aims to begin a process of considering life history differences between species in a wider range of ecological models to increase their accuracy and predictive power. The final chapter applies these predictions on life history and community dynamics to a well-studied and heavily censused tree community at Barro Colorado Island, Panama. I construct life history models for each of the 90 most abundant tree species on the island using a mix of adult and sapling censuses, and theory on reproductive allometry in trees. I then utilize these life history models to predict short-term fluctuations in population abundance using minimal parameters. I compare this to previous modeling approaches and demonstrate that, applying the theory of life history variation in ecological communities, models using only census observations can predict short-term dynamics as well as those with fitted, estimated parameters
Coverage and exploration planning with semantically informed maps
Get PDFSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-20 without embargo termsThe student, Joao Marcos Correia Marques, accepted the attached license on 2025-06-26 at 22:05.The student, Joao Marcos Correia Marques, submitted this Dissertation for approval on 2025-06-26 at 22:22.This Dissertation was approved for publication on 2025-07-02 at 14:53.DSpace SAF Submission Ingestion Package generated from Vireo submission #22373 on 2025-10-20 at 16:57:15Outside of controlled factory environments, robots are expected to build their own environment representations to effectively plan and act in the world. This representation typically takes the shape of a semantic map constructed through some Simultaneous Localization and Mapping pipeline. In this thesis, I first present my work on a robotic application in the wild whose performance relies not just on the accuracy of the resulting metric-semantic world map, but also on its uncertainty: Semantically Informed Ultraviolet Disinfection Planning. Taking this application as an example of semantically informed coverage planning, I then present evidence that most of the sequential estimation strategies for semantic maps lead to overconfident label estimates, and that this overconfidence impacts the downstream performance of the robotic agents, both in disinfection and object goal navigation tasks. I also provide two novel confidence-preserving memory-efficient methods to perform online metric-semantic reconstruction in real time. I then extend the domain of coverage planning problems to include the coverage of interactive scenes by proposing a solution to the Manipulation-Enhanced Mapping problem, where a robot must efficiently survey an environment with a camera and manipulate objects in the environment to improve object visibility. This is achieved by leveraging neural networks to accelerate the belief updates of well-studied formalism for decision-making under uncertainty of Partially Observable Markov Decision Processes, enabling us to solve these problems in the belief space of metric-semantic maps. I finally conclude with a summary of the presented research, as well as promising future research directions in the areas of coverage planning, interactive scene representations and on further improving confidence calibration and uncertainty handling in 3D metric-semantic maps
Self-assembly of anisotropic colloids
No full textSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-20 without embargo termsThe student, Bahadir Rusen Argun, accepted the attached license on 2025-07-09 at 14:35.The student, Bahadir Rusen Argun, submitted this Dissertation for approval on 2025-07-09 at 15:43.This Dissertation was approved for publication on 2025-07-11 at 15:58.DSpace SAF Submission Ingestion Package generated from Vireo submission #22463 on 2025-10-20 at 16:57:34Classical particle-based simulations—such as molecular dynamics and Monte Carlo methods—are typically conducted using spherical particles. However, the particles in many systems of interest are non-spherical, either by design, as in colloidal self-assembly, or as a result of natural processes, as observed in environmental nanoplastics. While efficient overlap detection algorithms enable the simulation of anisotropic shapes, they are generally limited to hard-core repulsive interactions. An alternative is the composite bead approach, which allows for the simulation of arbitrary particle shapes with both attractive and repulsive interactions. In this method, smaller spherical beads are rigidly connected to maintain the overall geometry of a larger, complex particle. This strategy offers flexibility, as it can represent a wide range of convex and concave shapes. We use this approach to model different nanoplastic particle morphologies. To evaluate their ecological impact, it is essential to understand the fate of nanoplastics in the environment. These particles are often surrounded by natural colloids, which can promote aggregation via favorable interactions. We perform molecular dynamics and multiparticle collision dynamics simulations to understand the effect of particle shape and flow on the heteroaggregate structure and breaking behavior. We find that mostly round particles formed compact structures with a large number of neighbors, weak connection strength, and a higher fractal dimension. Microplastics with sharper edges and corners aggregated into more fractal structures with fewer neighbors, but with stronger connections. We investigated the behavior of aggregates under shear flow. The critical shear rate at which the aggregates broke up is much larger for spherical and rounded cube microplastics. For these shapes, the compact aggregate structure outweighs their weaker connection strength. The rounded cube aggregate exhibited unexpectedly high resistance to breakup under shear. We attribute this to being fairly compact due to weaker, flexible neighbor connections, which are still strong enough to prevent particles from breaking off during shear flow. Irrespective of the stronger connections between neighboring microplastics, the fractal aggregates of cubes break up at lower shear rates. We find that cube aggregates reduced their radius of gyration significantly, indicating restructuring during shear, while most neighbor connections were kept intact. Sphere aggregates, however, kept their overall size while undergoing local rearrangements, breaking a significant portion of their neighbor interactions. To accurately represent the particle shapes and obtain smooth, realistic effective pair interactions between two rigid bodies, each body may need to contain hundreds of spherical beads. Given an interacting pair of particles, traditional molecular dynamics methods calculate all inter-body distances between the beads of the rigid bodies within a certain distance. For a system containing many anisotropic particles, these distance calculations are computationally costly and limit the attainable system size and simulation time. However, the effective interaction between two rigid particles should only depend on the distance between their center of masses and their relative orientation. Therefore, a function capable of directly mapping the center of mass distance and relative orientation to the interaction energy between the two rigid bodies, would completely bypass inter-bead distance calculations. It is challenging to derive such a general function analytically for almost any non-spherical rigid body. We have employed various machine learning tools to achieve this task. The pair configuration (center of mass distance and relative orientation) is taken as input and the energy, forces and torques between two rigid particles are predicted directly. We show that molecular dynamics simulations of cubes and cylinders performed with forces and torques obtained from the gradients of the energy neural-nets quantitatively match traditional simulations that use composite rigid bodies. Both structural quantities and dynamic measures are in agreement, while achieving up to 23 times speed up over traditional molecular dynamics, depending on hardware and system size. In addition to using shape-based descriptors, we also represented pair configurations through point groups that share the same invariances as the interacting rigid bodies. This formulation enabled efficient treatment of the symmetries inherent in the pair configurations and made it possible to leverage machine learning potentials originally developed for atoms and quantum mechanical datasets. We compared machine learning potentials based on predefined versus learnable descriptors and show that, although models with learnable descriptors can achieve high predictive accuracy, their architectural complexity leads to slower inference times, limiting their practical applicability. In contrast, Neuroevolution Potential descriptors combined with fully connected neural networks strike a favorable balance between accuracy and computational efficiency. Point-based descriptors coupled with full-connected neural networks exhibit better generalization across different particle geometries and are easier to implement. Overall, our results demonstrate that machine learning can substantially accelerate molecular dynamics simulations of anisotropic particles while accurately capturing their equilibrium structures