Lights out: Traffic controller will help cars pass intersections without stopping
CLEMSON, South Carolina – In the not-too-distant future, when all the cars on the roads drive themselves, traffic and traffic lights won’t be necessary. Americans, who currently spend 7 billion hours and 3 billion gallons of extra fuel a year sitting in traffic, will have to find new excuses for being late.
“Our cities are very congested and a lot of that congestion happens at traffic lights,” said Ardalan Vahidi, an associate professor of mechanical engineering in the College of Engineering, Computing and Applied Science at Clemson University. “If we can reduce that congestion and harmonize the flow of traffic through traffic lights, we can contribute to more efficient, cleaner cities where driving will also be more comfortable.”
Ali Reza Fayazi, who recently graduated from Clemson University with a doctoral degree in mechanical engineering, has created a way to facilitate that harmony and he’s tested it on a closed track where real cars interacted with simulated cars.
Autonomous cars of the future promise smoother traffic flows by communicating with each other, but cars talking amongst themselves won’t eliminate traffic problems. For that, cars and intersections also have to talk with each other, said Vahidi, an engineer who came to the United States from Iran in 1998 on a student visa. He became a U.S. citizen in 2013.
Fayazi’s intelligent traffic control system lives on a remote server and tracks all the cars approaching an intersection. It messages the cars with the best times to arrive and the cars use that data to adjust their speeds and paths so they can safely pass with as few stops as possible.
“Basically, vehicles subscribe to a server that manages an intersection,” said Vahidi.
Vehicles send GPS locations and velocities to the controller. The controller uses the data to solve a dynamic optimization problem and assigns each vehicle a time to arrive at the intersection based on safety and minimized stopping time.
Most traffic lights change at fixed times without considering upcoming traffic. A simulation of a typical intersection showed as many as 1,100 cars stopping in one hour.
In a simulation with the intelligent intersection controller, only 11 cars stopped.
After favorable simulation testing, Fayazi tested his traffic controller on a closed test track at the International Transportation Innovation Center (ITIC) in Greenville. He used his personal vehicle as a stand-in for an autonomous vehicle and the data the intelligent traffic controller would feed directly to an autonomous car’s computer instead fed into a smartphone app he created.
In a recent demonstration on the ITIC track, Fayazi and Vahidi each drove their personal cars through an imaginary intersection marked by orange traffic cones. A monitor in Fayazi’s car showed the real cars among simulated cars driving through the intersection.
A smartphone app in each car connected to the remote intelligent intersection controller via the cellular network. The controller tracked the cars and sent speed recommendations to the drivers. With the controller and the apps running the algorithms continuously, they were able to make real-time adjustments.
Testing the system on the track enabled Fayazi to advance the theory of having no traffic lights into reality, something that hasn’t been done before.
“Without a closed test track it is not possible to evaluate technologies like this, because real cars have to interact with simulated cars,” Fayazi said.
The ITIC track also enabled Fayazi to accurately measure fuel consumption and other data through another app he created that communicates with the vehicle’s on-board diagnostics port. When a real test vehicle interacted with the intelligent traffic controller at one intersection, it used 19 percent less fuel compared to another test with a typical intersection. That savings would increase with the number of intelligent intersections. Emissions would drop as well, Fayazi said, although that data was not measured.
“Right now, we hope we can use this technology in a smart city project where only autonomous vehicles are allowed to travel,” Fayazi said. With modifications, the traffic controller could be used with mixed-traffic, where person-driven cars share the road with autonomous cars, he said. Physical traffic lights would still be necessary, but person-driven cars could receive speed recommendations that would reduce stopping while providing safe distances between cars.
The next phase of research, Vahidi said, is connecting multiple intersections.
“An effective passage of platoons (of cars) requires real-time coordination of neighboring intersection controllers that can be quite challenging,” he said. “However, if intersection controllers share the same network, it would make communication of decisions very efficient.”
Research could also focus on the human experience inside a fully autonomous car that doesn’t stop at intersection without traffic lights, Fayazi said. “Will the passengers feel tense or fearful facing the conflicting stream of cars?”
The research supporting this work was funded by the National Science Foundation.