How to Beat the Traffic

Supercomputer that makes 8 million-billion calculations per second takes on highway congestion

- Damien Martin

Of all the aspects of pre-pandemic life people missed, sitting in traffic was not one. If you’ve returned to the office full- or part-time, you might have noticed it creeping back into your life. Traffic jams are not only annoying — the National Renewable Energy Laboratory calculated that each of us on average spends 40-50 hours in them each year — they waste a lot of energy, about 3.3 billion gallons of fuel annually. And all that time spent staring at taillights adds up to 8.8 billion hours and $10 billion of lost productivity per year. NREL wants to do something about that, so it put Eagle, a supercomputer that can make 8 million-billion calculations per second, to the task.

For a number of reasons — it’s not far from the Department of Energy’s Oak Ridge National Laboratory, it’s in the top 20 traffic-congested cities in the nation, and it’s a medium-sized metro area — NREL chose Chattanooga, Tenn., as its test case. NREL scientists created a digital twin of Chattanooga’s traffic patterns and set Eagle to work optimizing flow. The time- and gas-saving techniques it examined could be used in cities across the U.S. to cut down on commutes and reduce fuel consumption by as much as 20%.

Optimized Traffic Lights

The quickest fix NREL and Eagle found, using machine learning and data from GPS and vehicle sensors, was that traffic lights in a high-traffic area were timed for rush hour and not adjusted during off-peak times. That resulted in large chunks of the day when cars in lighter traffic were stopping at red lights for no real reason.