You're driving down a crowded highway and see an upcoming wave of glowing red brake lights. The car in front of you starts to brake, so you brake and cause the car behind you to do the same. Gridlock. A few miles down the road, when the logjam clears, you find that there was... nothing. No accident, no police pullover, no bottleneck, no highway construction. It was just pointless traffic with seemingly no cause. 

Traffic engineers call this frustrating phenomenon "a phantom traffic jam." It's different than aggressive, bumper-to-bumper rush hour traffic, which is an overloaded roadway problem. Rather, a phantom jam is something that naturally arises when people drive fast enough and close enough together. But the key word here is "people." It turns out that if even a tiny percentage of the cars on the road are autonomous—think of a highway dotted with just 2 percent of something like Google's self-driving car—then the number of phantom traffic jams drops drastically.

"Small perturbations in otherwise perfectly normal driving are gradually amplified by other cars."

Small effects add up

According to Benjamin Seibold, a mathematician at Temple University who specializes in traffic engineering simulations, a phantom traffic jam is not actually caused by bad drivers or mistakes made on the road. It's that small effects accelerate through the system. "What happens is that when a roadway becomes densely packed with vehicles, small perturbations in otherwise perfectly normal driving"—say, just barely letting off the gas and slowly by a couple of miles per hour—"are gradually amplified by other cars, until they actually become waves of cars suddenly braking."

So phantom traffic is typically no single person's fault. But that doesn't mean it's unavoidable. What would happen if self-driving cars, which don't make the same kinds of small errors that humans do, take to the highways?

Last year Seibold and a team of researchers received a three-year, $1 million grant from the National Science Foundation to find out what effect, if any, autonomous vehicles could have in alleviating traffic such as phantom traffic jams. Although Seibold is still in the very early stages of this project, he's put together computer models that demonstrate ways in which self-driving vehicles could alleviate and actively combat phantom traffic jams. Even a tiny spattering of such autonomous cars dispersed throughout a highway could act like vehicular jetties—breaking up the waves of slowing cars that would otherwise gradually congest into annoying traffic jams.

In one of Seibold's preliminary traffic models, he found that if autonomous cars accounted for just 2 percent of the traffic on the road, those robot cars would "drive in a particular way that makes them better at keeping a constant velocity, can reduce stop-and-go traffic by as much as 50 percent."

How a robot drives

How can so few vehicles make such a vast difference? Seibold says the self-driving cars would leave slightly more room between them and any vehicle ahead of them, braking and accelerating with carefully calculated smoothness. That is, instead of tailgating the cars in front of them and having to slam on the brakes all the time, like certain lead-footed human drivers, self-driving cars leave enough space to slow gently when things tighten up, creating fewer hard braking "waves" that spread through the traffic system.

Autonomous vehicles would also have an ability humans don't to sense traffic patterns up ahead. Assuming a self-driving car could sense what's happening up to 8 cars ahead, it could reduce an existing phantom traffic jam by preventatively slowing down just a few mph in advance of a block-up. Despite the fact that these aren't very big changes in driving style, these tiny corrections have a huge effect.

To be sure, Seibold is careful to note that his new models are still rather simple—ignoring factors like lane changes and the occasional rogue, terrible driver—and that live experiments using real, running autonomous vehicles are planned but have yet to be done. Nonetheless, he says, all evidence so far points towards the fact "these vehicles could really have a measurable impact on phantom traffic, even shortly after they start to enter the roadway."

Headshot of William Herkewitz
William Herkewitz
Science & Technology Reporter
William Herkewitz is a science and technology journalist based in Berlin, Germany. He writes about theoretical physics, AI, astronomy, board games, brewing and everything in between.