Microsoft is helping researchers in Brazil study whether they can predict traffic congestion 15 minutes to an hour before it happens.
The plan is to use Microsoft’s cloud computing service to store and crunch data from multiple sources.
“The Traffic Prediction Project uses data available by social networks, department of transportation, and data that the users create themselves while they move around the city,” Juliana Salles of Microsoft Research said in a promotional video.
Those data points aren’t new, and neither is work to predict traffic jams. In fact, such predictions are happening now, says Peter Keen of Digital Traffic Systems.
“It is being done now. It can always be done better,” Keen says, because right now prediction models depend on past traffic information.
“You have historical trends of what the volume’s going to be at a given day of the week, at a given time of day,” says Keen, which can help make predictions that are accurate much of the time, especially about typical traffic patterns on major passageways.
Keen says predictions can be better if both past and current data is combined. But such data is often in multiple formats and multiple databases, and hard to combine.
Rahim Benekohal, professor of civil and environmental engineering at the University of Illinois who studies traffic flows, says we already know a lot about traffic patterns.
“By understanding how the congestion grows, where the congestions grows and what’s the cause of it,” says Benekohal, you can often predict future congestion without the need for crunching huge amounts of data.
In fact, Benekohal says, most of us can predict traffic congestion just from our own past experience with around 80 percent accuracy. Still, he says he admires the effort being undertaken by Brazilian researchers with the help of Microsoft.
Keen adds that there have been a lot of efforts to research traffic congestion prediction, but no one has yet discovered “a magic bullet,” he says.