In 2024, the Texas A&M Institute of Data Science (TAMIDS) created the Seed Program for AI, Computing, and Data Science (SPAICD) to fund projects that enhance novel ideas, foster innovation, initiate new collaborations, and develop results to support further funding opportunities within the fields of data science, artificial intelligence, and machine learning.
In the 2024 cycle, TAMIDS received 39 proposals from faculty across the University. A panel reviewed each proposal and scored against a rubric expressing the SPAICD program’s goals and criteria. Out of these 39 submissions, we awarded the top ten reviewed proposals, spanning twelve departments in five colleges, schools, and divisions, totaling 22 funded researchers. These projects cover various topics, including autonomous vehicles, personalized medicine, and environmental science, exemplifying TAMIDS’s mission to support innovative research projects and foster new multidisciplinary partnerships. Below are the awarded proposals and researchers.
2024 Funded Proposals
Leveraging Deep Learning and Big Data Analytics to Enhance Tropical Cyclone Seasonal Predictions with Spatial Heterogeneity
Tropical cyclones (TCs) are among the deadliest and most destructive climate extremes, impacting millions each year. While predictions of individual TC track and intensity on short timescales have significantly improved, progress in long-lead seasonal prediction remains limited. Unlike stochastic short-term weather predictions, TC seasonal prediction is fundamentally a data science problem, with current skills hindered by a lack of high-quality data and reliance on simplistic statistical methods. To address these challenges, we propose developing an innovative seasonal TC prediction system leveraging deep learning techniques and big data analytics. This system would skillfully predict two-dimensional TC spatial distributions in the summer season as early as winter. We aim to develop more complex prediction systems that could significantly enhance US TC prediction capabilities and attract interest from the private sector, with the developed methodology also serving as a basis for extramural funding applications.
Keywords: tropical cyclone; climate seasonal predictions; geographical spatial heterogeneity; generative artificial intelligence;
Researchers
- Dan Fu, Assistant Professor, Atmospheric Sciences
- Zhengzhong Tu, Assistant Professor, Computer Science & Engineering
AIMS-META: AI-enhanced designing of Manufacturability-aware and Symmetry-driven METAmaterials for enhanced mechanical performance
Lattice metamaterials with exceptional mechanical properties present significant potential for diverse applications. This proposed research seeks to address key challenges in the design and fabrication of these materials through a comprehensive approach. By employing mathematical frameworks and advanced computational techniques, the study aims to explore novel design spaces that could uncover unprecedented possibilities. Machine learning models are proposed to further refine the generation and optimization of designs, ensuring both innovation and manufacturability. Additionally, integrating real-time monitoring systems during fabrication is expected to enhance precision and reduce defects. The goal is to establish a systematic methodology for advancing material design and broadening its practical applicability across various domains.
Keywords: Mechanical metamaterials, group symmetry, 3D printing, periodic structures
Researchers
- Suparno Bhattacharyya, Texas A&M Institute of Data Science
- Samuel Gonumakulapalle Lodi, Industrial & Systems Engineering
- Satish Bukkapatnam, Industrial & Systems Engineering
Optimizing Urban Vehicle Fleet Through Multi-Agent Reinforcement Learning: Integrating Built Environment Factors and Real-Time Road States
This project aims to develop a multi-agent Reinforcement Learning (RL) framework utilizing Centralized Training with Decentralized Execution (CTDE) to optimize fleet sizes in urban areas, focusing on autonomous vehicle deployment. As cities face increasing traffic congestion and air pollution challenges, the need for efficient and sustainable transportation solutions is critical. The proposed RL framework integrates dynamic urban factors to create a realistic environment that reflects the complexities of urban mobility. The framework will incorporate multi-source data such as traffic state data and sensor data to provide real-time updates, ensuring the model adapts dynamically to changing urban conditions. This research not only advances the understanding of RL applications in transportation but also aims to set the foundation for smarter and more sustainable cities. The anticipated outcomes include improved air quality, reduced traffic congestion, and enhanced mobility, directly contributing to sustainable urban development.
Keywords: Fleet Optimization, Multi-agent Reinforcement Learning, Built Environment, Autonomous Vehicle
Researchers
- Xinyu Li, Department of Landscape Architecture & Urban Planning
- Suphanut Jamonnak, Texas A&M Institute of Data Science
Enhancing Privacy and Security in Machine Learning and Data Science
The proposal addresses the security and privacy aspects of Machine Learning and Data Science tools in a variety of cloud and edge cloud scenarios. It is motivated by significant recent interest in distributed architectures in which different units are responsible for data collection, processing, and storage. For example, the data collected by a security camera can be processed at the cloud edge and archived at a cloud data center. While decoupled cloud-based architectures have many important advantages, they can also create additional threat vectors and expand the surface area for potential attacks. In particular, significant privacy and security concerns are associated with remote processing of sensitive data.
Keywords: Security, Privacy, Neural Group Testing
Researchers
- Alex Sprintson, Department of Electrical & Computer Engineering
- Krishna Narayanan, Department of Electrical and Computer Engineering
Mapping Cool Corridors: Thermal Exposure and Shade Dynamics for a Walkable TAMU
This proposal aims to generate high-resolution thermal exposure and shade change maps of the Texas A&M campus to guide the development of “cool corridors.” Cool corridors are pathways designed to minimize thermal exposure by maximizing natural shade or incorporating artificial shading structures. This study will create a digital model of building surfaces and tree canopies, which will serve as inputs for shadow function in the microclimate model SOlar and LongWave Environmental Irradiance Geometry model (SOLWEIG) to produce hourly shade distribution maps for the campus. These shadow maps will be overlaid with the campus sidewalk map through an overlay analysis, resulting in hourly sidewalk shadow area maps. By integrating these shadow maps and campus human mobility, the project will provide actionable insights for campus planners to design more walkable and comfortable cool corridors.
Keywords: TAMU Campus, Shade coverage, cool corridor
Researchers
- Cuiling Liu, Department of Landscape Architecture & Urban Planning
- Yapping-Xu, Department of Environmental and Geosciences, Sam Houston State University
Leveraging AI to Enhance the Urban Planning Process: Detecting Opportunities in Comprehensive and General Plans
This project seeks to improve the urban planning process by applying advanced artificial intelligence (AI), data science (DS), and machine learning (ML) techniques to the analysis of comprehensive and general plans from diverse U.S. communities, including urban, suburban, and rural areas. These plans are essential for long-term development but often rely on only traditional methods. By employing AI-driven techniques like predictive modeling, pattern recognition, and optimization, the study aims to enhance data-driven decision-making, public engagement, and scenario analysis in planning. The interdisciplinary research team, consisting of experts from urban planning, geography, and computer science, will analyze more than 1,000 planning documents (as a pilot study) to identify recurring themes and innovative methodologies. The project aligns with Texas A&M’s strategic goals in advancing smart city technologies and AI innovation. Collaborations with the American Planning Association will help extend the project’s impact and visibility.
Keywords: comprehensive plans, planning, AI, text mining
Researchers
- Thomas Sanchez, Department of Landscape Architecture & Urban Planning
- Lei Zou, Department of Geography
- James Caverlee, Department of Computer Science & Engineering
AutoFLUKA: An AI-Assisted Framework for Automating Monte Carlo Simulations
Monte Carlo simulations are critical tools for modeling complex physical interactions across various scientific and engineering disciplines, including high-energy physics, medical physics, nuclear engineering, and environmental science. However, performing Monte Carlo simulations for intricate and advanced scenarios presents significant technical challenges and a steep learning curve. Maintaining Monte Carlo algorithms necessitates dedicated research and development efforts to establish toolsets that accelerate education and training in this area. Our proposal, termed AutoFLUKA, seeks to address this critical need by developing foundational methods to alleviate the steep learning curve associated with Monte Carlo modeling, thereby enhancing automation and integration capabilities. AutoFLUKA will be an AI-assisted framework that integrates Large Language Models (LLMs) with domain-specific knowledge to automate and streamline Monte Carlo simulation workflows. By employing Retrieval-Augmented Generation (RAG) techniques, AutoFLUKA can provide intelligent assistance in modifying input files, executing simulations, and processing results for visualization. This approach not only reduces human labor and errors but also accelerates research and development processes.
Keywords: Monte Carlo Simulations, FLUKA, Large Language Model, Retrieval-Augmented-Generation
Researchers
- Yang Liu, Department of Nuclear Engineering
- Jian Tao, Department of Visualization
Cross-Domain Knowledge Transfer for Extrapolation in Natural Hazards Engineering
The long-term vision of this project is to leverage currently available data to better understand the performance of built infrastructure under extreme hazardous loads that will inevitably arise due to climate change, paving the way for a new convergent research area in data-driven extrapolation for extreme scenarios. Currently, data-driven approaches are exceptional at making inferences and decisions, however, existing data-driven or artificial intelligence/machine learning techniques are limited in their ability to extrapolate and make reliable predictions for unprecedented extreme events or conditions not represented in the training data. Developing robust data-driven models capable of reliable extrapolation under such dynamic, non-stationary conditions with limited training data poses significant technical hurdles. In this project, we will compose a highly cross-disciplinary framework for data-driven extrapolation in natural hazards engineering. The framework will leverage transfer learning, a popular small-data approach, and submodular optimization for selecting optimal data subsets for knowledge transfer.
Keywords: Data-driven extrapolation; knowledge transfer; optimization; algorithms
Researchers
- Victoria Crawford, Department of Civil & Environmental Engineering
- Stephanie Paal, Department of Civil & Environmental Engineering
An Integrated Digital Twin and Computer Vision Model for a Proactive Work Zone Intrusion Warning System
Highway construction and maintenance activities pose significant safety risks to both drivers and workers, resulting in a high number of fatalities and injuries. Although work zone intrusion alert systems have gained significant attention, current systems are reactive, frequently generate false alarms, and do not give workers and drivers sufficient time to avoid accidents. To address the challenges of current practices, the proposed research aims to (1) develop a novel proactive computer vision and digital twin-based approach that predicts work zone intrusions by detecting vehicles from a great distance, estimating their speed, and tracking their movement, allowing drivers to react promptly to alerts and avoid entering a work zone and (2) validate the performance of the developed detection algorithm in a controlled but real-world environment. This study will contribute to promoting the adoption of AI in safety-related fields by demonstrating how digital twin technology can generate synthetic data and enhance the performance of AI models.
Keywords: Digital Twin, Computer Vision, Work Zone Safety, Work Zone Intrusion
Researchers
- Namgyun Kim, Department of Construction Science
- Yalong Pi, Texas A&M Institute of Data Science
Toward Smart Orthopedic Surgery Planning by using Physics-Informed Machine Learning
Dental implants enhance quality of life, but their failure can be critical. Success depends on the strain levels applied to teeth, which are influenced by bone stiffness. Since stiffness varies across bone geometry, precise 3D point-wise estimation and strain prediction are crucial. However, current methods are either invasive, limited to macro-level data, or too slow for clinical use. This proposal aims to enhance dental surgery planning through novel physics-informed machine learning methods to provide accurate, rapid bone stiffness inference and strain prediction. By integrating CT data, clinical factors, and biomechanical principles with advanced machine learning, the project seeks to deliver personalized bone assessments for surgery planning and predict bone behavior during surgery. This interdisciplinary approach, combining data science, biomechanics, and healthcare, is expected to advance dental surgery planning, reduce implant failures, and extend to orthopedic surgery and other medical fields. The project aligns with national and institutional priorities.
Keywords: Personalized Dental Surgery Planning, Bone Heterogeneous Elasticity Estimation, Physics-Informed Machine Learning, Personalized Medicine
Researchers
- Jaesung Lee, Department of Industrial & Systems Engineering
- Yuxiao Zhou, Department of Mechanical Engineering