Project Title

Investigating Relationship between Driving Patterns and Traffic Safety using Smartphones Based Mobile Sensor Data

Collaborating Universities

Old Dominion University
1 Old Dominion University
Norfolk, VA 23529

Principal Investigator(s)

Rajesh Paleti (ODU) Email:
Mecit Cetin (ODU) Email:

Funding Source(s) and Amounts Provided (by each agency or organization)

$71,967 (VDOT)

Total Project Costs

$71,967 (State)

Start Date


Completion Date



In spite of various advancements in vehicle safety technologies and improved roadway design practices, roadway crashes remain a major challenge. While certain hotspots may be unsafe primarily due to the geometric features of these locations, in many cases the safety risk seems to be an outcome of the unsafe driving patterns along the roadway stretching downstream and/or upstream of the actual crash locations. Even though there is plenty of research on correlating safety measures to roadway characteristics and some elements of traffic flow (e.g., exposure, speed), there is no significant literature on analyzing the correlation between high-resolution speed and acceleration data and crash risks along highway segments. Collecting such high-resolution data is now feasible with the mobile consumer devices such as smartphones. Smartphones are now equipped with sensors capable of recording vehicle performance data at a very fine temporal resolution in a cost-effective way. The current project used this mobile sensor data to identify unsafe driving patterns and quantified the relationship between these driving patterns and traffic crash incidences.


This study will use innovative data collection methods and develop improved statistical models. So, this research is expected to result in good journal publications. Also, the study findings will be presented in national conferences focused on data collection in transportation including North American Travel Monitoring Exposition and Conference (NATMEC) in Miami, FL and Transportation Research Board Meeting in DC.


  1. Data from mobile sensors (e.g., smartphones) enables monitoring vehicle dynamics and traffic flow at high resolution (e.g., second-by-second). This data can be used to develop microscopic traffic measures that serve as better indicators of actual driving patterns. There are opportunities to expand the mobile sensor data collection on a larger scale (e.g., statewide or nationwide) by partnering with probe data providers (e.g., INRIX, HERE). Currently, most of the Safety Performance Functions (SPFs) are only sensitive to aggregate variables such as traffic exposure and geometric attributes (e.g.: presence of shoulder). Federal safety agencies and state DOTs would benefit immensely by updating these SPFs using microscopic traffic measures so that the predictions are more accurate and these models can also be used to evaluate countermeasures that primarily affect driving behavior (e.g., variable speed limits).
  2. Currently, most of the SPFs in the HSM are either Poisson or Negative Binomial models. However, there is considerable scope for improving these models without adding significant computational complexity. Specifically, the Heterogeneous Dispersion (HD) and Generalized Ordered Response (GOR) variants of the NB model are relatively easy to estimate and were found to improve the statistical fit significantly. To start-off, state DOTs can test and evaluate the relative merits (better prediction accuracy and policy sensitive) and difficulties (data collection) of these models for specific locations (e.g., freeways, intersections etc.).

Web Links to Reports and to the Project website