UTSA Runner Research Press (Univ. of Texas at San Antonio)
Not a member yet
6846 research outputs found
Sort by
On the Global Practical Exponential Stability of h-Manifolds for Impulsive Reaction–Diffusion Cohen–Grossberg Neural Networks with Time-Varying Delays
No full textIn this paper, we focus on <i>h</i>-manifolds related to impulsive reaction&ndash;diffusion Cohen&ndash;Grossberg neural networks with time-varying delays. By constructing a new Lyapunov-type function and a comparison principle, sufficient conditions that guarantee the global practical exponential stability of specific states are established. The states of interest are determined by the so-called <i>h</i>-manifolds, i.e., manifolds defined by a specific function <i>h</i>, which is essential for various applied problems in imposing constraints on their dynamics. The established criteria are less restrictive for the variable domain and diffusion coefficients. The effect of some uncertain parameters on the stability behavior is also considered and a robust practical stability analysis is proposed. In addition, the obtained <i>h</i>-manifolds&rsquo; practical stability results are applied to a bidirectional associative memory (BAM) neural network model with impulsive perturbations and time-varying delays. Appropriate examples are discussed.Mathematic
Risk-Based Multi-Threat Decision-Support Methodology for Long-Term Bridge Asset Management-Volume 1: AI-Based Bridge-Level Decision Support
No full textThis project develops a methodology and tools for a risk-based multi-threat decision-support tool for long-term bridge asset management (BAM), with a particular focus on chronic aging-induced condition deterioration, and more abrupt and extreme seismic hazard impact. Specifically, a stochastic bridge condition deterioration and seismic damage simulation module is developed. Bridge condition deterioration is modeled through Markovian state transition dynamics considering different maintenance actions. Seismic fragility modeling and risk assessment is carried out, considering site-specific seismic hazard and the effect of seismic retrofitting actions. A life cycle cost analysis module is introduced to holistically quantify and aggregate the direct and indirect costs incurred from bridge condition deterioration, seismic damage, and intervention actions over a planning horizon. Benefit-cost analysis for various seismic retrofitting actions is also performed. Finally, by integrating the above bridge deterioration and seismic damage simulation module and the life-cycle cost analysis module with the advanced AI technique, deep reinforcement learning (DRL), a methodology for generating AI-based policies for sequential maintenance decision support for a portfolio of bridges is proposed. Departing from traditional condition-based decision policies, these AI-based policies can offer much more proactive and adaptive decisions to minimize the long-term life-cycle costs. Owing to the parametrized DRL formulation, the AI-based policies can flexibly accommodate the decision needs from different individual bridges within a bridge portfolio in near real time. Practical action constraints are also introduced to align with real-world engineering practices. The proposed AI-based policies are evaluated based on individual bridges as well as on a portfolio of bridges, and demonstrate superior performance in reducing the life-cycle costs compared with other condition-based policies. In addition, the AI-based policies also exhibit robustness to potential human override. Finally, effect of seismic retrofitting, when coupled with AI-based agents, is evaluated for more comprehensive life-cycle benefit-cost evaluation of seismic retrofit actions.Federal Highway Administration (FHWA)Civil and Environmental Engineering, and Construction Managemen
A Comparison of the Effects of Model Adaptation Techniques on Large Language Models for Non-Linguistic and Linguistic Tasks
Get PDFGenerative large language models (LLMs) have revolutionized natural language processing (NLP) by demonstrating exceptional performance in interpreting and generating human language. There has been some exploration of their application to non-linguistic tasks, which could lead to significant advancements in fields that rely heavily on structured data and specialized knowledge. However, there has been limited direct comparison of the effects of model adaptation techniques for non-linguistic compared to linguistic tasks with LLMs. To this end, the work in this paper investigates the effects of fine-tuning and few-shot learning on pre-trained LLMs for non-linguistic tasks using chess puzzles as a case study task. We compare the impact of fine-tuning and few-shot learning on models performing the same task represented in both chess notation (i.e., non-linguistic data) and natural language descriptions of the same chess notations (i.e., natural language data). Our experiments with Mixtral-8x7B-v0.1 and Meta-Llama-3-70B resulted in a 5% lower average increase in performance after fine-tuning for non-linguistic tasks compared to linguistic tasks. Similarly, few-shot learning on pre-trained models exhibited a 3% lower average increase in performance for on non-linguistic tasks compared to linguistic tasks. Furthermore, few-shot learning on fine-tuned models resulted in a significant accuracy drop, particularly for Mixtral, with a 24.82% decrease for non-linguistic tasks. These results suggest that fine-tuning and few-shot learning for generative LLMs have stronger effects on linguistic tasks and their data than for non-linguistic.Computer Scienc
A Group-Based Participatory Approach to Examine Community Resilience and Trauma
Get PDFBackground
The St Louis Resiliency in Communities After Stress and Trauma (ReCAST) project promoted community well-being in a designated Promise Zone over a 5-year period. The primary goals of the ReCAST project are: (1) to build a foundation to promote well-being, resiliency, and community healing through service integration; (2) to improve access to trauma-informed community behavioral health resources and youth peer support; and (3) to create community change through community and youth engagement, leadership development, improved governance, and capacity building.
Purpose
To examine ReCAST stakeholder perceptions of resilience and trauma in their communities using Concept Mapping (CM), a participatory mixed methodology.
Approach
CM is an integrated approach that supports the structured conceptualization of ideas and applies multidimensional scaling and hierarchical cluster analysis to bring together and organize the ideas of a group to capture the perspectives of multiple stakeholders.
Results
33 stakeholders participated in the ReCAST program. The resilience concept map yielded five clusters: (1) Community Relationships (positive and productive), (2) Religious Organizations/Spirituality, (3) Interaction with Diverse Communities, (4) People in power to create change, and (5) Community gatherings and organizations. The trauma cluster map identified the following clusters: (1) Substandard Education, (2) Traumatic events/community violence, (3) Racial Trauma, and (4) Physical Degradation of neighborhoods.
Discussion
Based on CM results, participants identified the need for local political officials to make a coordinated effort to address issues expressed in both maps and funding of local organizations to address issues related to trauma and resilience, especially youth organizations.Public Healt
COMMUNICATE, UNDERSTAND, AND COLLABORATE: AN AI-INTEGRATED HUMAN-ROBOT COLLABORATION APPROACH TO IMPROVE INTUITIVENESS AND PRODUCTIVITY IN CONSTRUCTION
No full textThe full text of this item is not available at this time because the author has placed this item under an embargo until December 11, 2027.Despite critical safety and productivity problems that the construction sector faces, a growing labor shortage further exacerbates these issues. The U.S. construction market size reached trillions of dollars in 2023 and is expected to have an average annual growth rate of 5.3% from 2025 to 2027. However, 91% of contractors reported difficulties in recruiting skilled workers, and these worker shortages are expected to remain elevated (Associated Builders and Contractors (ABC), 2024) Such expanding challenges could impede market growth and aggravate safety and productivity issues greatly. To address this problem, adopting advanced technology, such as robotics, is a promising solution to mitigate the impact of labor shortages.
Considering the complexity of construction tasks (e.g., various tools and materials for different stages) and dynamic working environments (e.g., vehicles and workers coexisting on a site), autonomous robotics are introduced as assistants to complete tasks collaboratively with humans, a concept known as Human-Robot Collaboration (HRC). Existing studies have shown that collaborative robots can assist in physically demanding construction tasks. However, construction workers are often reluctant to use them, leading to HRC implementation failures. Most current HRC approaches require humans to supervise and plan for robots (e.g., using joysticks or pre-programming). This necessitates intensive training for workers and imposes additional cognitive loads, adding extra burdens on them. Moreover, these approaches do not resemble natural user interfaces – the intuitive interactions we have when collaborating with another human worker – which affects workers' trust and acceptance of HRC.
The overarching goal of this research is to achieve proactive and intuitive HRC in construction by improving robots' intelligent capabilities to understand and adapt to workers’ operation needs and establish transparent communication between humans and robots. Towards this goal, three fundamental knowledge gaps have been identified, and each of them is addressed in a specific objective in this research.
The first knowledge gap is the lack of an analysis on the impact of physical HRC implementation on work performance and worker perceptions in construction. This research applies an experimental study to reveal the practical impact of HRC on work productivity and worker perceptions. By comparing the physical HRC and conventional Human-Human Collaboration (HHC) on the same construction task, workers' actual perceptions and attitudes towards the robot and their work performance are investigated. These findings will enrich our understanding of HRC implementation in construction activities and obtain guidance on possible improvements we could make for a more intuitive HRC. The expected outcome is the quantification of impacts of physical HRC on work performance and worker perceptions in construction tasks, and users’ attitudes/expectations on future HRC approaches.
The second knowledge gap is the lack of a method that predicts human motions considering both humans and robots in construction tasks. Human behavior is purposeful and influenced by working environment contexts, especially robot movements in HRC. This research proposes a context-aware deep learning approach to integrate different context information into 3D human motion prediction models. Various context integrations are validated and examined for more accurate predictions in 3D human motion prediction. A table-top experiment with ten participants is conducted to collect human movements, robot trajectories, and assigned targets locations. A two-branch deep learning framework is introduced to enhance the accuracy of 3D human motion prediction. The contextual information, including full-robot motions, motion of only robot’s gripper, and object locations, is investigated in three models. The model comparison illustrates the significance of contextual information in human motion predictions for safe and effective HRC.
The third knowledge gap is the lack of intuitive and transparent (bidirectional communication) HRC approach in construction. This research proposes an Artificial Intelligence (AI) integrated robotic system for workers to interact with robots via natural user interfaces (e.g., speech and gaze), and for robots to reason and plan the work itself as well as provide audio feedback. The bidirectional communication between humans and robots is established by taking in human speech and gaze in a large-language model (LLM) and providing robot feedback (e.g., ask for more information or provide its plans). This system is evaluated and implemented in a demonstration. The expected outcome is an AI-integrated robotic system that improves the transparency and approach of HRC in construction. It further improves the trust of workers in collaborative robots to accelerate the successful HRC implementation in construction.Civil and Environmental Engineerin
RELATIONSHIPS AMONG SPACE WEATHER ELEMENTS AND THEIR IMPORTANCE IN FORECASTING
No full textThe full text of this item is not available at this time because the author has placed this item under an embargo until December 11, 2026.Understanding and forecasting space weather is essential for protecting technological systems and human activities in space. Solar activity such as flares, coronal mass ejections (CMEs), interplanetary shocks, solar energetic particles (SEPs), and energetic storm particle (ESP) events can disrupt communications, damage spacecraft, and pose radiation hazards to astronauts and aviation. This dissertation investigates these processes and their role in driving space weather impacts, with a focus on SEP forecasting.
A statistical study of interplanetary shocks and ESPs examines how sampling windows influence derived properties, providing constraints for modeling and forecasting. Chapter 3 addresses the characteristics and hazards of SEPs, which can reach Earth within minutes and pose severe radiation risks. To improve prediction, a machine learning framework—the Multivariate Ensemble of Models for Probabilistic Forecast of SEPs (MEMPSEP)—was developed. Using convolutional neural networks in an ensemble approach, MEMPSEP provides probabilistic SEP occurrence forecasts and regression-based predictions of onset time, peak flux, and event duration.
Reliable machine learning models rely on robust, multimodal datasets. Here, we describe the construction of these datasets for SEP forecasting and assess how variations in dataset composition affect MEMPSEP performance. We further use Permutation Feature Importance (PFI) to quantify the contributions of individual features.
Overall, this work advances the understanding of SEP drivers, demonstrates the utility of machine learning for probabilistic and regression-based forecasting, and emphasizes the critical role of data quality in developing operational space weather prediction tools.Physics and Astronom
RACE, RESOURCES, AND CHILD WELFARE: A COMPARATIVE STUDY OF THE PERCEPTIONS OF CHILD ABUSE AND NEGLECT PREVENTION FUNDING IN TWO TEXAS CITIES
Get PDFChild abuse and neglect is a critical issue millions of families face across the United States. Child abuse has negative implications like long-term chronic illness and adult substance abuse affecting families, children and society; however, the influence of race is often ignored when disusing those implications. The purpose of this study is to understand how stakeholders perceive the distribution of child abuse and neglect prevention funding, particularly for counseling services in San Antonio and Austin and in what ways this distribution affects access to resources in predominantly minority communities compared to predominantly white communities. This study used semi-structured interviews of local stakeholders in Texas and community health reports from San Antonio and Austin Texas to find themes that suggest there is a perceived difference in prevention resource access and distribution between certain racial demographics.Political Scienc
Toward Real-Time Measurements of Toxic and Flammable Gases Using Wavelength Modulation Spectroscopy
Get PDFA series of sensitive measurement techniques based on laser absorption spectroscopy of toxic and flammable gases in harsh environments is presented, with applications towards portable sensing for first-responders and firefighters. Scanned-wavelength direct absorption spectroscopy and wavelength-modulation spectroscopy methods for detection of hydrogen chloride, hydrogen cyanide, methane, and acetylene are developed and applied in progressively harsh and chaotic reacting flow environments; spanning application from barometrically-prepared mixtures of the target gases in well-characterized optical cells, to sampled effluent gases generated by pyrolysis of solid fuels in a laboratory fume hood bench-scale reactor, and ultimately to sampled gases collected from full-scale compartment fires wherein multiple synthetic materials were simultaneously burned to represent a residential or commercial fire. The measurement techniques are in aggregate demonstrated to perform sensitive measurements below 100 ppm of the target gases with a compact sensor form-factor, laying the groundwork for future technology developers and fire researchers to (1) downscale the photonics and optics towards the commercial development of a compact portable sensor for fire personnel, and (2) perform laboratory studies aimed at characterizing the fundamental pyrolysis and combustion behavior of solid fuels. With respect to (1), the wavelength modulation spectroscopy technique is then further developed towards a real-time measurement paradigm, using multiple mixtures of flammable gases as a test bed for the development and refinement of a multi-variable linear regression-based technique to report species concentrations at 1 Hz time resolution.Mechanical Engineerin
A Novel Approach to State-to-State Transformation in Quantum Computing
Get PDFThis article presents a new approach to the problem of transforming one quantum state into another. It is shown that an <inline-formula><math display="inline"><semantics><mrow><mi>r</mi></mrow></semantics></math></inline-formula>-qubit superposition <inline-formula><math display="inline"><semantics><mrow><mo stretchy="false">|</mo><mrow><mi mathvariant="bold-italic">x</mi></mrow><mo stretchy="false">&#10217;</mo></mrow></semantics></math></inline-formula> can be obtained from another <inline-formula><math display="inline"><semantics><mrow><mi>r</mi></mrow></semantics></math></inline-formula>-qubit superposition <inline-formula><math display="inline"><semantics><mrow><mo stretchy="false">|</mo><mrow><mi mathvariant="bold-italic">y</mi></mrow><mo stretchy="false">&#10217;</mo></mrow></semantics></math></inline-formula>, by using only <inline-formula><math display="inline"><semantics><mrow><msup><mrow><mo>(</mo><mn>2</mn></mrow><mrow><mi>r</mi></mrow></msup><mo>&minus;</mo><mn>1</mn><mo>)</mo></mrow></semantics></math></inline-formula> rotations, each presented by one controlled rotation gate. The quantum superpositions with real amplitudes are considered. The traditional two-stage approach <inline-formula><math display="inline"><semantics><mrow><msubsup><mrow><mi>U</mi></mrow><mrow><mi>y</mi></mrow><mrow><mo>&minus;</mo><mn>1</mn></mrow></msubsup><msub><mrow><mi>U</mi></mrow><mrow><mi>x</mi></mrow></msub><mo>:</mo><mo stretchy="false">|</mo><mrow><mi mathvariant="bold-italic">x</mi></mrow><mo stretchy="false">&#10217;</mo><mo>&rarr;</mo><mo stretchy="false">|</mo><msup><mrow><mrow><mn>0</mn></mrow><mo stretchy="false">&#10217;</mo></mrow><mrow><mo>&oplus;</mo><mi>r</mi></mrow></msup><mo>&rarr;</mo><mo stretchy="false">|</mo><mrow><mi mathvariant="bold-italic">y</mi></mrow><mo stretchy="false">&#10217;</mo></mrow></semantics></math></inline-formula> requires twice as many rotations. Here, both transformations to the conventual basis state, <inline-formula><math display="inline"><semantics><mrow><msub><mrow><mi>U</mi></mrow><mrow><mi>x</mi></mrow></msub><mo>:</mo><mo>&nbsp;</mo><mo stretchy="false">|</mo><mrow><mi mathvariant="bold-italic">x</mi></mrow><mo stretchy="false">&#10217;</mo><mo>&rarr;</mo><mo>&nbsp;</mo><mo stretchy="false">|</mo><msup><mrow><mrow><mn>0</mn></mrow><mo stretchy="false">&#10217;</mo></mrow><mrow><mo>&oplus;</mo><mi>r</mi></mrow></msup></mrow></semantics></math></inline-formula> and <inline-formula><math display="inline"><semantics><mrow><msub><mrow><mi>U</mi></mrow><mrow><mi>y</mi></mrow></msub><mo>:</mo><mo>&nbsp;</mo><mo stretchy="false">|</mo><mrow><mi mathvariant="bold-italic">y</mi></mrow><mo stretchy="false">&#10217;</mo><mo stretchy="false">&rarr;</mo><mo>&nbsp;</mo><mo stretchy="false">|</mo><msup><mrow><mrow><mn>0</mn></mrow><mo stretchy="false">&#10217;</mo></mrow><mrow><mo>&oplus;</mo><mi>r</mi></mrow></msup></mrow></semantics></math></inline-formula>, use <inline-formula><math display="inline"><semantics><mrow><msup><mrow><mo>(</mo><mn>2</mn></mrow><mrow><mi>r</mi></mrow></msup><mo>&minus;</mo><mn>1</mn><mo>)</mo></mrow></semantics></math></inline-formula> rotations each on two binary planes, and many of these rotations require additional sets of CNOTs to be represented as 1- or 2-qubit-controlled gates. The proposed method is based on the concept of the discrete signal-induced heap transform (DsiHT) which is unitary and generated by a vector and a set of angular equations with given parameters. The quantum analog of this transform is described. The main characteristic of the DsiHT is the path of processing the data. It is shown that there exist such fast paths that allow for effective computing of the DsiHT, which leads to the simple quantum circuits for state preparation and transformation. Examples of such paths are given and quantum circuits for preparation and transformation of 2-, 3-, and 4-qubits are described in detail. CNOT gates are not used, but only controlled gates of elementary rotations around the <inline-formula><math display="inline"><semantics><mrow><mi>y</mi></mrow></semantics></math></inline-formula>-axis. It is shown that the transformation and, in particular, only rotation gates with control qubits are required for initialization of 2-, 3-, and 4-qubits. The quantum circuits are simple and have a recursive form, which makes them easy to implement for arbitrary <inline-formula><math display="inline"><semantics><mrow><mi>r</mi></mrow></semantics></math></inline-formula>-qubit superposition, with <inline-formula><math display="inline"><semantics><mrow><mi>r</mi><mo>&ge;</mo><mn>2</mn><mo>.</mo></mrow></semantics></math></inline-formula> This approach significantly reduces the complexity of quantum state transformations, paving the way for more efficient quantum algorithms and practical implementations on near-term quantum devices.Electrical and Computer Engineerin
Design, Development, and Analysis of an Edge Machine Learning Andon System for Safe Human-Robot Interaction
Get PDFThis thesis presents the design, development, and analysis of the Human Aware Andon Module (HAAM), a low-cost, energy-efficient edge machine learning system for enhancing safety in human-robot interactions in manufacturing and industrial environments. The system integrates a Coral Dev Board Micro with an Edge TPU, a Time-of-Flight (ToF) sensor, and RGB LEDs to detect humans, estimate distances, and visually signal robot states. A novel depth estimation algorithm is developed that aligns low-resolution ToF data with RGB camera input. Using the aligned depth data and detection bounding boxes, a weighted cell-based distance calculation is used to provide a distance estimate for each human detection in a scene. We demonstrate the configurability of the HAAM system through integration testing with a UR5e robot in a pick-and-place scenario. Testing showed successful speed modulation and emergency stopping with response times under 10ms, providing real-world validation of the system's capabilities as a practical safety solution. The HAAM system offers significant advantages in cost ($104), power efficiency (1.27W), and visual signaling capabilities compared to commercial alternatives, demonstrating the potential of edge machine learning for enhancing workplace safety through visual feedback of machine states. More so, the small form factor and low power requirements allow for the HAAM to be easily attached to a variety of robot platforms.Mechanical Engineerin