亚洲AV

Big data may lead to safer roadways, lower emissions

In This Story

People Mentioned in This Story
Body

Once, transportation officials made decisions based on household surveys performed roughly once per decade, which asked selected households to record their travel behavior on a given day. With the advent of smartphones, similar data became available roughly every few minutes. Now, with an increasing number of connected vehicles on the road, that data is available in nearly real time. CEIE professor Shanjiang Zhu is embracing this shift, exploring the capabilities of researchers with this massive amount of data.聽

Shanjiang Zhu

With funding from the Virginia Department of Transportation (VDOT), Zhu and his research team, Anand N. Vidyashankar from the department of statistics and Chenfeng Xiong from the Civil and Environmental Engineering department at Villanova University, will reconcile the travel data from three different sources鈥攕urveys, smartphones, and connected vehicles鈥攊nto invaluable travel information.聽

"In the past, we tried to understand travel behavior, which is critical for future investment decisions and also transportation policy, based on survey data,鈥 Zhu explained. 鈥淏ased on that, you understand, on average, where people have traveled, in what mode, with whom, and spent how much time there, uh what is the purpose for the trip, etc. Using that information, you can develop a model that basically can predict future scenario, like how congested the network could be in 2040; and that drives all the investment decisions and policy debates.鈥 This method introduces problems of timeliness, as it can skip major events such as the COVID-19 pandemic, and human error, as people would not necessarily remember every detail of their travel on a given day.聽

The introduction of widespread smartphone use about ten years ago made the available data much denser, said Zhu, resulting in about one data point every three to five minutes. Each time a person鈥檚 smartphone app calls for location service, their location is automatically registered. Nevertheless, this method introduced a bias problem, as not everyone owns a smartphone and not everyone uses them often. 聽

Drone image of highway traffic at night

About a year ago, Zhu鈥檚 team won a competition hosted by VDOT to make the best possible use of connected vehicle data, basically newer vehicles like those with an 鈥淪OS鈥 button installed. One drawback to the data currently is connected vehicles currently make up a relatively small share of vehicles on the road.聽

鈥淏ut we have ways to make corrections from a statistical perspective, and then this gives you a much more accurate picture of traffic on the road,鈥 said Zhu, adding 鈥淥n average, it's one data point every three seconds. With such data, the accuracy and timeliness of travel demand models could be greatly improved.鈥 Zhu noted his colleague Vidyashankar will be reviewing the data fusion to ensure a rigorous statistical approach.聽

The new data also opens the door for new safety studies, Zhu said, adding safety studies are currently based mainly on police reports after an accident has occurred. By using alternate data, such as how often a car鈥檚 brake deceleration rate exceeds a certain threshold or how hard a driver turns the steering wheel, dangerous locations might be addressed before an accident occurs. Zhu is interested in exploring the topic further using the dataset resulting from his current project.聽聽聽

Zhu foresees data from connected vehicles becoming increasingly important as more and more people adopt the technology. He said, 鈥淣ow we are investing in the methodology part and seeing how we can make this connection more productive, to improve the driving environment, to make our roads safer, to make the driving experience better, and also to reduce our energy consumption and emissions."聽