Transforming Urban Mobility Policy Through Accessible AI-Driven Framework

• In Event: Small Area & Geographic Estimation with New Data and New Methods

Abstract

Traffic congestion is a persistent public policy challenge affecting economic productivity, environmental sustainability, and quality of life in urban areas. As cities grapple with growing populations and infrastructure limitations, traditional engineering congestion mitigation approaches – such as road expansions and static traffic signals – have proven both costly and insufficiently adaptive to dynamic urban environments. This study proposes a novel, policy-relevant framework that uses geospatial Artificial Intelligence (GeoAI) and predictive analytics to inform data-driven transportation policy and infrastructure planning.

Using 15 years of traffic data from New York City, this research applies time-series forecasting models (ARIMA, LSTM) and geospatial clustering techniques within Uber’s H3 Hexagonal Grid framework to identify and predict congestion patterns at a hyperlocal level. By integrating these insights with a custom-built AI-powered query tool based on the LLaMA large language model (LLM), the study enables a real-time, interactive toolset of traffic data for urban planners and policymakers. This tool supports natural language queries, allowing decision-makers to dynamically assess congestion risks, evaluate infrastructure performance, and simulate potential policy interventions.

The study responds directly to policy imperatives outlined by urban sustainability and decarbonization strategies, presenting AI not merely as a technical solution but as a strategic policy instrument. It demonstrates how smart mobility data can be operationalized to shift from reactive congestion responses to proactive, equity-informed urban transport policies, addressing both systemic inefficiencies and environmental goals. By advancing predictive spatial analytics, this research contributes actionable knowledge to the broader policy discourse on smart cities, climate resilience, and sustainable transportation planning. Its methodology and tools are scalable, offering a replicable model for other municipalities aiming to integrate AI-driven insights into public infrastructure governance and evidence-based policy development.

Author

• Anton Rozhkov, New York University
  Presenting Author