Skip to main content

Incident Response

In collaboration with TDOT, Vanderbilt researchers designed a data-driven forecasting framework that can aid the spatial distribution of safety patrol vehicles on highways. By leveraging the power of big data and machine learning, the approach designed at Vanderbilt is based on combining terabytes of data related to roadway geometry, weather, historical accidents, and traffic. State-of-the-art machine learning techniques are used to extract patterns from massive data that can estimate the likelihood of incident occurrence across highway segments spread over 100,000 sq. km in Tennessee. The predictive model is subsequently used to proactively station safety vehicles in anticipation of future incidents. Initial results demonstrate that response times for patrol vehicles can be reduced by close to 20% on average. Understanding the spatial-temporal dynamics of accident occurrence and the strategic allocation of resources can optimize and significantly improve the safety of the highways.

Such statistical procedures are being developed in general to help communities across the globe. These are important because First-responders are constrained by limited resources, and must attend to different types of incidents like traffic accidents, fires, and distress calls while considering the dynamically changing environments under which critical emergency response occurs. An overview of the tool is available at  


Example Projects:

The work on a Review of Incident Prediction, Resource Allocation, and Dispatch Models for Emergency Management is available here.

An exploratory work on understanding the theoretical aspects of response procedures was published here. A new version of this dashboard is available here

A similar study for Tennessee department of transportation was undertaken with specific challenges of high sparsity and high resolution incident forecasting. The dashboard showing the results is available here.



Hiba Baroud

Abhishek Dubey

Ayan Mukhopadhyay