By Kevin Wilcox
A research team finds that if just a small percentage of vehicles were automated, the beneficial effect on traffic flow could be huge.
The cars traveled around a circle until the wave pattern appeared. The automated car was then switched to the computer algorithm, and the waves dissipated. Courtesy of John de Dios and Alan Davis
May 23, 2017—Every motorist knows the frustration of a traffic jam, cars inching along a highway wasting time and fuel. Many don't realize, however, that some of these traffic jams are phantoms, caused by small variations in the speed of just one driver that then are amplified by the actions of other drivers.
One of the promises of automated vehicles is that they will eliminate such effects, which can be compared to amplified oscillation waves that ripple through traffic, often prompting hard-braking maneuvers that can cause rear-end collisions, which in turn cause further congestion. New research indicates that this solution may be much closer than many people think.
A research team comprising multiple institutions and funded by a grant from the National Science Foundation recently published resultsof a series of experiments carried out in the Arizona desert in which a single automated car traveling in a circle with 20 or 21 driver-controlled vehicles successfully eliminated the waves that can cause traffic jams. The tests validated earlier work by the team that used sophisticated computer modeling.
The experiments were "an adventure," says Benedetto Piccoli, Ph.D., the associate provost for research at Rutgers University, who points out that the logistical challenges included assembling a fleet of 25 vehicles, finding a site suitable for a circular path, and coordinating the many facets of the experiment. "You need to instruct drivers," Piccoli says. "The cars need to start at a given position each time. You need to run the experiment for some time. And on top of that, there was the heat of the Arizona desert that rendered everything even more interesting."
Piccoli is one of the principal investigators in the research, along with Benjamin Seibold, Ph.D., an associate professor of mathematics at Temple University; Jonathan Sprinkle, Ph.D., the Litton Industries John M. Leonis Distinguished Associate Professor in the electrical and computer engineering department at the University of Arizona at Tucson; and Daniel B. Work, Ph.D., an assistant professor in the civil and environmental engineering department at the University of Illinois at Urbana-Champaign.
Three experiments were conducted in Tucson, where the intense summer sun forced the team to begin preparations at 4 AM. At the start of each test, each vehicle was separated precisely from the one behind it by the same distance on a large circular circuit on an expanse of pavement. At the beginning of each test, a human driver in the autonomous vehicle had full control of the car. When researchers observed the waves appearing, they switched to the autonomous system, which was controlled by an algorithm they had developed in computer testing.
"Similarly, at the end of the experiment, we were switching back off the autonomous part, and we let the driver on board drive normally," Piccoli says. "And we saw the waves reappear again. So, the experiment was: Waves appearing, destroying waves with our autonomous driving, and then switching back to human driving and the waves reappearing again.
"The outcomes were great," Piccoli says. "Indeed, we show that using appropriate control algorithms, this single autonomous vehicle was able to destroy the waves and to improve on various measures—including braking events and vehicle fuel consumption—to the point that it was even surprising to us."
In what the researchers referred to as experiment A, braking events were reduced 98.6 percent, and fuel consumption was reduced a staggering 42.5 percent. In experiment B, braking events were reduced by 76.2 percent and fuel consumption by 22.1 percent. In experiment C, braking events were reduced by 74.4 percent and fuel consumption by 28.1 percent.
"Personally, I was very surprised at the magnitude of the fuel consumption benefits," said Work, who provided written answers to questions posed by
"Considering that only one vehicle in the experiment was automated, it is quite remarkable that the control had such a large benefit [over] the other human-driven vehicles."
Piccoli attributes the strong effect to the combination of the autonomous car not creating any of its own oscillations and absorbing the oscillations of human drivers by not overreacting. This had a calming effect on all the cars in the experiment. "Sometimes you can underreact while still keeping a safe distance," Piccoli explains. "By taking this action of reacting appropriately or even underreacting, [the autonomous car] was able to absorb the oscillations created by the human drivers before they would reach a stage of generating a wave."
The experiment simulated a single lane of traffic. The researchers are now planning similar research involving a multilane highway simulation. In multilane settings, lane changes have a pronounced effect on traffic flow. "The multilane is probably the biggest issue in moving this research to the next stage of impact," Piccoli says. "And that's exactly what we're tackling now." The team is considering scenarios with a single autonomous vehicle or multiple ones, he adds.
In the meantime, Piccoli says that although fully autonomous vehicles aren't likely to constitute 5 percent of the traffic volume on public roads—as was the case in the experiment—anytime soon, smart technologies in the form of computer-aided braking and cruise control systems that gauge the speed of the car ahead and make fluid adjustments are already entering the market and have the potential to have a huge effect.