Whether you’re in Boston, Berlin or Beijing, traffic feels the same. Getting stuck in a traffic jam is frustrating and makes you feel helpless. Sometimes it happens unexpectedly and sometimes there are no really good alternative options, so you’re just stuck there. It often hits when you are most in a hurry and, despite your best intuition, it feels tough to predict.
Even very modern technology solutions can easily be surprised by how traffic builds up and delays occur. One problem is that much of it looks at how the traffic is moving in particular spots right now. Which turns out to be an only-partially reliable guide to how the traffic will be when your car arrives.
As Strategy Analytics pointed out a couple of weeks ago, “This approach to traffic reporting is not unlike the weatherman looking out the window – or looking at historical weather reports. We all know that that is not how forecasting is done.”
Traffic prediction is a very different problem to traffic reporting, and a much harder nut to crack.
The reality is that traffic congestion is a hyper-local problem (it depends on the particular set-up of roads and usage demands in each city), which is why it is often so hard to predict.
On most days, on most stretches of the road in the world, there are actually no tail-backs. Instead, congestion hits in targeted spurts at very particular times of the day, week or year in very particular stretches of road going in a certain direction.
In the afternoon of the fourth Wednesday of November, for example, roads near major American airports and big thoroughfares are likely to be packed with people trying to get to their families for Thanksgiving. But downtown streets might be deserted as everyone takes off early. In Canada and Germany, however, the fourth Wednesday in November is just like any other November Wednesday.
“Traffic is a highly nuanced problem,” says Tony Belkin, HERE Director of Traffic. “There are dozens of different variables from weather to holidays to events that can affect how traffic is moving on a given stretch of road. You have to take them all into account if you want to properly analyze and understand how traffic is moving.”
What feels like common sense about traffic - that a road is packed on the weekdays but empty on the weekends, is often far more complex.
Two years in traffic
Tony and his HERE traffic team have spent nearly two years studying the problem and trying to tease out every variable they could think of that affects traffic. They profiled traffic behavior and identified over 100 different ways that traffic can behave on a given stretch of road.
Then the next stage was to identify which of those factors apply at any point. As part of this work, the team developed an algorithm to identify the appropriate profile for any given moment over any stretch of road.
It's June and this road leads towards the beach. It's mid-morning Saturday... historically, this has led to some heavy traffic. The system selects a profile that describes this combination.
Grinding down roads
Tony’s team worked round the clock to analyze every major roadway in the United States, Canada and Germany, crunching billions of data points in order to fine tune predictions for how traffic would behave at any given moment.
“Our goal is to crack the code of traffic predictions so that drivers could more accurately figure out how long it would take them to get to their destination,” said Tony. “They could then decide that, hey, maybe it makes more sense to wait 15 minutes before leaving so the traffic will clear or to jump in the car now to beat the expected jam.”
And after two years, they managed to crack that code. If you want to meet us and learn more about how we approach traffic, you can learn more about predictive traffic at Mondial, check out the HERE booth at the Paris Motor Show at Pavilion 3, Booth 240.