AI-Driven Digital Twins for Transportation Environments: Integrating Geospatial Scene Synthesis with VOC Exposure and Short-Term Health Risk Modeling

Abstract

The operational complexity of urban transportation systems increasingly requires computational environments that can represent physical infrastructure, dynamic mobility flows, and environmental stressors in an integrated manner. Digital twins have emerged as promising platforms for coupling sensing, simulation, and decision support, yet many existing implementations remain fragmented across geometric reconstruction, air quality assessment, and public health analytics. This paper presents a system-level analysis of AI-driven digital twins for transportation environments, emphasizing the integration of geospatial scene synthesis with volatile organic compound exposure and short-term health risk modeling. Rather than proposing a single algorithm, the paper examines architectural trade-offs, data infrastructure requirements, model governance mechanisms, robustness conditions, fairness considerations, and deployment constraints. The discussion develops a layered conceptual architecture in which generative geospatial scene synthesis supplies microenvironmental context, while exposure models translate pollutant concentrations into health-relevant indicators across transit modes. Particular attention is given to the challenges of synchronizing heterogeneous data streams, managing uncertainty in AI-generated scene representations, and ensuring that health risk outputs remain interpretable for public agencies. The paper further addresses policy implications related to air quality management, transit ventilation improvements, and the equitable distribution of environmental information. It argues that the value of such digital twins lies not primarily in predictive precision alone, but in their capacity to link spatial representation, environmental monitoring, and health governance within a transparent and maintainable socio-technical system.