Just a few self driving cars on the road could improve traffic for everyone

BERKELEY, Calif. (KGO) -- It happens to all of us. You are driving along the freeway when, for no obvious reason, some cars start slowing down while others are still cruising along. It is what experts call a "phantom" traffic jam.

Vanderbilt University traffic researcher Dan Work says you are getting stuck simply because of the way people drive.

Work explains that "small disturbances from one driver get magnified by the driver behind. Pretty soon it speeds up, slow down, speed up, slow down."

Work knows all about the phantom traffic jam. His team even ran an elaborate field test, demonstrating it is a real phenomenon.

Video of the test shows 22 cars driving in a circle. Researchers told the drivers to drive as they would normally in rush hour traffic. As the cars keep circling, the video shows they eventually get into a stop and go pattern. The cars race around one side of the track, only to get caught in the traffic jam on the other side of the track.

The next part of the experiment shows a promising solution. The same drivers keep circling, but this time there is a self-driving car, with artificial intelligence, mixed into the group. The car is programmed to smooth the flow of traffic. The human drivers don't know it is there, but they still respond to its steady speed. So as the self-driving car moves more smoothly, all the other cars do too.

"What the self-driving cars have to do is basically drive at the average speed of that person who is driving too fast and then too slow," Work said. A human driver cannot figure out that magic traffic-calming speed, but a car with artificial intelligence can.

Here in the Bay Area, traffic researchers at the Department of Energy's Lawrence Berkeley National Laboratory are using sophisticated computer modeling to predict what else self-driving cars can do to improve traffic.

The project is led by Alex Bayen, director of U.C. Berkeley's Institute of Transportation Studies. Bayen showed us several computer simulations of traffic situations with cars programmed to behave like human drivers, mixed with a self-driving car with artificial intelligence.

The computer models show the self-driving cars figure out on their own how to improve efficiency when cars are merging on to a road or moving through an uncontrolled intersection.

The team is also trying to figure out whether automated cars could smooth out traffic flowing on to the San Francisco-Oakland Bay Bridge. Bayen says the cars could communicate with each other and possibly regulate the speed of traffic approaching the toll plaza, maybe even replacing metering lights.

"In the coming years, we will see an increasing proportion of vehicles on the road with some level of automation, not fully automated, but some level of automation. That means within a few years we will have maybe five percent or ten percent of vehicles that could help control traffic," Bayen said.

Another big benefit of smoothing traffic flow with automated vehicles is energy savings. Work reports his team's research, mixing regular human-driven vehicles with one self-driving car in a controlled setting, reduced fuel consumption by 40 percent. Work says even though the numbers on actual roads might be somewhat lower, that is an amazing result.

"If you think about a 40 percent fuel reduction, not just of the self-driving car, but all the other cars that are in the traffic flow, it really shows the potential that these automation systems, even before every car has them, can benefit all of our commutes," according to Work.

Work says many of the best selling cars in the United States already offer an early version of this type of technology as at least an optional feature. The systems use an advanced version of cruise control that allows you to set the amount of space between your car and the one ahead. Researchers say that can help keep you driving more smoothly - and benefit everyone around you.