We commonly hear promises and praises regarding automated cars: that they will eliminate human-error accidents and make roadways and traffic more efficient. At first glance, such promises seem to indicate that self-driving cars would reduce congestion and reduce our commute times.
But researchers aren’t so certain.
Hesham Rakha, an engineering professor at Virginia Tech, specializes in traffic’s flow — or lack of flow.
One of the basic challenges of traffic, he explains, is that it’s extremely easy for human drivers to cause traffic slowdowns. “One little touching of the brake can pull the whole system down,” he says.
That’s primarily because of delayed human reactions. If one driver spots red tail lights and taps her own brakes in response, all the drivers behind her have to react as well, and that delay forces every person to brake harder than the next. In almost no time the line of cars it at a standstill for practically no reason.
Even the smallest disruption can create a 10 to 20% reduction in flow, Rakha says. “If we can prevent that congestion from happening, not only will we let vehicles travel faster, but we will also get more vehicles through our system.”
To show how automation might change traffic, a Planet Money journalist met Rakha at a parking structure overlooking the intersection of Interstate 66 and the Dulles Toll Road just outside Washington, D.C. That convergence zone is a notorious bottleneck where four lanes of traffic merge into two.
Rakha has already collected traffic data from sensors along the highway and created an algorithm to calculate how actual traffic might shift once more self-driving cars are in use.
One small tap of the brake can mess up the entire system for countless drivers.
Automation helps introduce real-time traffic information so cars know what’s ahead, can slow down to avoid disruptions, and anticipate the way the larger system is working so the funnel doesn’t get clogged. It also helps that “reaction time” isn’t an issue in an automated car.
Self-driving cars become synchronized. They communicate with other self-driving cars, alerting each other to disturbances ahead and adjusting to the optimal speed to avoid backup.
In Rakha’s simulation, as more self-driving cars hit the roads, the red dots representing congestion on the screen clear away and become green, moving dots.
Geoff Wardle, director of transportation at the ArtCenter College of Design, states that most experts expect automated cars to result in more ride- or car-sharing; also, they will park themselves. Either way, he says, “a significant amount of traffic congestion is caused by going around and around the same block looking for a place to park.”
But automation is not a cure-all for commuter’s headaches, either. The reason, ironically, once again comes back to human beings.
“We tend to like to try to cheat the system to gain personal advantage,” Wardle says. Drivers would be tempted to cut off driverless cars, once again creating inefficiencies in the trafffic flow.
In fact, Rakha’s research indicates that it could take a rather long time to finally improve traffic. That’s because while the behavior of automated cars is predictable, the actions of humans are not. So we likely won’t see a significant easing of congestion until a critical mass of driverless cars are finally on the road.
In the end, Rakha tells me he cannot predict whether automated cars will improve traffic congestion.
“I don’t know the answer,” he says. “If the road is less congested, more people are going to be attracted to that road, and so basically it will become congested because it’s supply and demand.”
If traffic does improve, more cars and trucks are likely to use the roads, and we could be back where we started.