In January, we presented our research at the Transportation Research Board (TRB) Annual Meeting during the poster session for the 2025 INRIX x MetroLab Challenge.

At Columbia University, our team is part of the Center for Smart Streetscapes, a National Science Foundation Engineering Research Center dedicated to advancing urban infrastructure through data, sensing, and analytics. 

Our project tackled a persistent challenge in urban mobility analytics: estimating representative traffic volumes across a citywide road network, including locations with limited sensing information. Leveraging INRIX datasets, we conducted a validation study of traffic volume estimation in New York City using ground truth traffic camera observations. Following this study, our team developed a traffic volume estimation model that predicts time-consistent traffic flow propagation across the citywide road network. 

Validating Traffic Volume at Scale 

To address this challenge, we leveraged datasets from INRIX including INRIX Trips and INRIX Speed Distribution Profiles. These datasets provided strong constraints for estimating citywide flows while preserving physically consistent traffic propagation across connected road segments. The main challenge was sparse and heterogeneous sampling (many segments had limited observations), which made calibration with traffic cameras essential for constructing a representative traffic model. 

What We Discovered 

Our first step was to compare INRIX-reported traffic volumes with ground-truth traffic camera observations across NYC corridors.  

We learned that:  

• The discrepancy between INRIX-reported average volumes and camera-based counts varied significantly by segment and time of day. 
• Trips trajectory signals often represented only a small fraction of total traffic. 
• Trajectory data alone was insufficient for producing representative citywide volume estimates. 

While trajectory data captures traffic movement patterns well, it must be aligned with physical observations to accurately reflect total flow. 

From Research to Real-World Impact 

Our study has immediate value for improving travel-time prediction for Emergency Medical Services, supporting faster routing and, by extension, patient care. In addition, the developed traffic model can be used for “what-if” scenario analysis enabling testing responses to incidents, special events, and street closures to inform planning, resource staging, and city policy decisions. 

The primary stakeholders include NYC agencies responsible for traffic management, traffic impacts of future development, along with emergency-response departments that rely on accurate, time-varying traffic volumes. 

Next Steps 

After presenting our work, we received strong feedback on its practical relevance and engaged in several follow-up discussions about potential deployments and partner use cases. The next steps include analyzing the impact of exogenous events (e.g., city parades, holidays, and weather conditions) on traffic volumes and flow propagation patterns and incorporating these factors into a traffic model that can anticipate likely scenarios. This information will equip traffic operations agencies with decision-support tools for planning, resource staging, and real-time response. In the future, our team is interested in extending the approach to multiple U.S. cities to assess transferability and move toward a scalable cyber-physical representation of urban road networks. 

A Step Toward Predictive, Network-Aware Cities 

By validating probe-based traffic data against ground-truth camera observations and embedding it within a network-consistent model, our work advances more reliable, citywide traffic intelligence. This framework supports improved emergency routing, operational planning, and scenario testing. Ultimately, it lays the groundwork for scalable, predictive urban mobility systems that help cities anticipate disruptions and manage transportation networks more effectively. 

Learn more about the 2026 INRIX x MetroLab Challenge.