As a Masters (’07) and Doctoral (’10) student at UC Berkeley, Daniel Work created a firm foundation on scholarship and innovation, a network of top scholars and researchers and a spirit of forging ahead with large-scale projects.
Now he’s transforming the way we think about how autonomous vehicles and emerging technology can revolutionize traffic control.
“The inspiration for our project grew out of a concept we developed at Berkeley,” says Work. “There are a lot of parallels with ground-breaking work of the Mobile Millennium project, where a small number of GPS equipped cars were used as sensors to monitor traffic. In our latest work, we followed the same concept with a small number of autonomous vehicles acting as actuators to control the traffic flow.”
In a recent multi-disciplinary study, funded by the National Science Foundation’s Cyber-Physical Systems program, Work, an assistant professor at the University of Illinois at Urbana-Champaign, is collaborating with researchers, also with Berkeley ties, to show that by adding just a few autonomous vehicles into a transportation ecosystem, they can eliminate stop-and-go traffic tendencies created by human drivers, reduce accident risk and increase fuel efficiency.
“A small number of autonomous/connected cars is not going to eliminate congestion, but it can help stabilize the traffic flow,” says Work. “The study shows there may be immediate benefits to include automated vehicles the traffic stream even at low penetration rates.”
The team of researchers with expertise in traffic flow theory, control theory, robotics, cyber-physical systems, and transportation engineering also include Joseph and Loretta Lopez Chair Professor of Mathematics at Rutgers University-Camden Benedetto Piccoli, who helped Work with traffic models used in UC Berkeley’s Mobile Millennium project, associate professor of Mathematics at Temple University Benjamin Seibold, who worked at Lawrence Berkeley Lab and UC Berkeley in 2000-2001, and Litton Industries John M. Leonis Distinguished Associate Professor in Electrical and Computer Engineering at the University of Arizona Jonathan Sprinkle, who ran the UC Berkeley self-driving car team working for College of Engineering Dean Shankar Sastry.
For the study, researchers placed one autonomous car on a track with 20 cars with human drivers and filmed their progress in Tucson, Ariz. When humans drive the cars, they create stop-and-go waves even in the absence of bottlenecks, lane changes, merges or other disruptions. With the introduction of an autonomous vehicle used to control the pace, researchers found a 98 percent decrease in extreme braking.
"Before we carried out these experiments, I did not know how straightforward it could be to positively affect the flow of traffic," Sprinkle said. "I assumed we would need sophisticated control techniques, but what we showed was that controllers which are staples of undergraduate control theory will do the trick."
Work says an unexpected outcome of the experiment was regarding fuel consumption. They saw a 40 percent reduction across all vehicles in the experiment when the autonomous vehicle actively controlled the flow. The test also looked at average speed in traffic flow with the introduction of an autonomous car, but they saw only a 10 percent increase in average speed.
While fully autonomous vehicles haven’t made their way into mainstream traffic yet, and connected autonomous vehicles are even further away, the research suggests that current technology, like adaptive cruise control, can make a difference now.
"Fully autonomous vehicles in common traffic may be still far away in the future due to many technological, market and policy constraints," Piccoli said. "However, increased communication among vehicles and increased levels of autonomy in human-driven vehicles is in the near future."
Researchers say they aren’t finished with their experiment. Next steps include looking at impacts of autonomous vehicles in denser traffic, where human drivers have more freedom, like changing lanes.
"The proper design of autonomous vehicles requires a profound understanding of the reaction of humans to them,” Seibold said, “and traffic experiments play a crucial role in understanding this interplay of human and robotic agents."
This material is based upon work supported by the National Science Foundation under Grant No. CNS-1446715 (B.P.), CNS-1446690 (B.S.), CNS-1446435 (J.S.), and CNS-1446702 (D.W.). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
Media credit: John de Dios, Alan Davis