Research
City Topology Through Graph Learning: Enhancing Urban Planning & Resilience
This research leverages Graph Neural Networks (GNNs) to analyze the topology of city infrastructure networks, including transportation, buildings, and land use. By learning the structural and functional relationships between these components, the study uncovers their role in shaping a city's economic and social performance under various stressors. The insights gained help inform policy and decision-making, enabling cities to enhance resilience, sustainability, and efficiency in urban planning and infrastructure management.
Deep Learning-Powered Sensor Fusion for Soil Moisture and Salinity Mapping
We have developed a novel sensor fusion technology that integrates acoustic impedance, electric capacitance, and conductivity measurements with a deep learning model to accurately assess moisture and salinity in soil. This innovation eliminates the need for frequent site-specific calibration by learning the complex interactions between moisture, salinity, and soil type.
AI-driven Scour Maintenance Strategy for Aging Bridge Systems in Flood-prone Zones
Deep reinforcement learning model was developed to optimize allocation of resources to mitigate scour-induced bridge failures, considering flood degradation, time deterioration, and community vulnerability. By simulating multiple flood scenarios using GIS data, the AI-driven approach outperforms traditional maintenance methods, offering cost-effective, data-driven solutions for sustainable infrastructure management.
Publication: [1]
Bayesian Network Modeling for Tracing PFAS Contamination
This study employs Bayesian network modeling to uncover the causal relationships between environmental (e.g., precipitation, elevation, land cover) and anthropogenic (e.g., factories, fire stations, airports) factors influencing PFAS contamination. The model not only predicts PFAS concentrations based on contributing variables but also infers potential contamination sources from observed PFAS levels, offering a powerful tool for environmental monitoring and pollution mitigation.