Predicting Highway Crashes
A new traffic model pinpoints where and when accidents happen, flagging particularly dangerous stretches of highway.
Researchers at Ohio State University have developed computer modeling software that identifies the probability of traffic accidents at certain times and locations on state roadways. The program–the first of its kind in the nation–is based on historical crash data. It uses existing statistical and mapping software to create a color-coded geographical display of the accident-risk levels on segments of roadway throughout the state.
“The model is saying, ‘This area has a higher risk than this area at this specific time of having this specific type of crash,’ which lets us predict where and when there are going to be higher risks of crashes,” says Christopher Holloman, the project leader and the associate director of the Statistical Consulting Service in Ohio State’s Department of Statistics. Currently, the model is being used by the Ohio State Highway Patrol to monitor roadways and position troopers. Eventually, the researchers would like to feed the data to drivers through mobile devices or portable navigation systems.
The predictive crash model was initially developed as a tool to help the Ohio State Highway Patrol better prevent accidents and explore the reasons some roads are riskier than others. Scientists have taken the historical crash data collected over the past five years by the highway patrol, which tracks details of accidents–including time, location, weather conditions, and whether alcohol or speeding was involved–and analyzed the data for roadway trends using statistical analysis software from SAS, the Cary, NC-based software giant. The software provides an output of the numerical risk levels for every piece of roadway. So on a particular day, one could look and learn which roadways have the highest risk of, for instance, alcohol-related crashes.
What makes this model novel is that scientists have now combined the statistical software with Google Earth–a program that offers an interactive map of the entire globe–to map the results as color-coded lines. Google Earth is able to perform this function because it reads the output from the statistical model in KML files; much as a Web browser reads HTML files, the KML files tell the program where on the planet to draw lines or place images, explains Holloman.
“We have done reports on individual places, on a specific weekend, to look at where the most dangerous spots are for people to watch out for,” he says. “We can make predictions for every major roadway in Ohio, under all possible road conditions, for every hour of the day, for every day of the week.”
“The main use for this type of technology, which is pretty straightforward, would be in the public sector: working with government and state departments of transportation to provide them [with] that information so they could make modifications, whether it is designs of the road or different signage, to protect drivers,” says Bryan Mistele, the founder and CEO of Inrix, a startup based in Kirkland, WA, that provides real-time and predictive traffic information.
Mistele finds the initial Ohio State concept intriguing. He says the next step for the scientists is to clearly indicate if higher-risk highways are correlated to the amount of traffic on the road because it is widely known that there is a significant correlation between the number of cars on the road and the propensity for accidents. By separating this information, researchers will have a better understanding of what makes certain roadways riskier than others and what type of accidents are prone to happen on these roadways.
Holloman’s group is continuing to work on the model to include more types of data and trends, such as where the police have been stationed and if low crash rates are related to the proximity of the highway patrol. The group also plans to study the reasons a certain roadway is more prone to alcohol-related crashes or having drivers who speed.