Rev, brake, honk, rev, brake. The pattern seems familiar to most of us. In 2015, American commuters spent more than 8 billion hours in traffic. If a robotics professor at Carnegie Mellon University has his way, that number could shrink.
Meet Stephen Smith, who has outfitted Pittsburgh with traffic signals that use artificial intelligence to react to traffic conditions in real time. These smart lights reduce travel time by 25 percent, braking by 30 percent and idling by more than 40 percent.
It began in Smith’s research lab at Carnegie Mellon. About six years ago, a local philanthropist with seed money wanted the university to address transportation in the Pittsburgh area. In 2012, Smith’s team installed a network of nine signals across the city.
“We’re fortunate that we have a really progressive thinking administration in the city. We have a great working relationship with the public works department there. They’ve been really helpful,” Smith said. “To be honest, they were a little hesitant to let us try at first, but with the backing of local foundations, they decided to give it a try. And then once they saw it was working, they became much more of a partner in the situation. The local administration here is really keen on putting technology to use and trying to become a model city of the future.”
Based on the positive results of the pilot, Smith obtained funding to expand the network. The smart signals now guide traffic at a network of 50 intersections in the east end of Pittsburgh.
In 2015, the company Rapid Flow Technologies spun out of Carnegie Mellon. The company, which develops intelligent transportation technology, expects to install its Surtrac traffic light system in about 150 more intersections in the next three years, covering about a third of Pittsburgh.
How It Works
Each traffic light makes its own decisions regarding when to turn red or green. First, Rapid Flow puts a computer at each intersection with a detection camera or radar.
“Each individual intersection watches the traffic approaching it and in real time, it builds a timing plan, a sort of plan of how much green time it’s going to give to each approach, so that the vehicles it’s seen through its detection get through as efficiently as possible,” said Smith, who serves as CEO of Rapid Flow.
After a smart signal studies the situation, it communicates expected traffic to its downstream neighbors. Those neighboring traffic lights go through the same process, each intersection working with information relayed by another intersection.
“It’s those intersections communicating with one another that get us this coordinated network behavior, progressions of green as you move,” Smith said.
Most smart traffic lights in the U.S. concentrate on main artery roads, but Rapid Flow focuses on grid networks that have “multiple dominant flows across traffic that can meet with one another and change through the day,” Smith said.
“In those kinds of settings, you really can’t count on the fact that you know what the dominant flow is. The system has to discover it at the time, and that’s what our system does.”
And it does it in real time, computing time plans in less than a subsecond. The signal’s goal is to compute as quickly as the traffic speeding through the intersection.
The traffic lights also hold the power of prediction.
“We can actually project out what the traffic is going to be like one or two or three cycles from now, a cycle being the full time through a green light,” Smith said. “That’s what allows us to communicate information between intersections and get this coordinating picture of what’s going to happen over the next two, three, four minutes.”
On the Road to the Future
Based on the reduced travel times in Pittsburgh’s smart intersections, Rapid Flow wants to bring the technology to cities across the U.S. It has entered negotiations with Los Angeles area municipalities, Burlington, Vermont, and eastern Pennsylvania.
Meanwhile, Rapid Flow has three full-time employees, not including Smith who works part time. The company expects to grow its staff over the next year.
Rapid Flow also has other projects in the works to make roads more efficient. During the past few years, the company has worked on making its real-time signal control work for connected cars equipped with radios.
“If two vehicles have these radios, they basically each put out a message every couple of seconds saying, ‘Here I am. Here’s my direction, and here’s my speed.’ And then if they get too close to each other, they can either alert the driver to take adverse evasive action or maybe just do it autonomously,” Smith said.
With Rapid Flow’s technology, intersections can tell these connected cars if the light is about to turn red or green, allowing the vehicle to adjust. Half of Rapid Flow’s 50 intersections in Pittsburgh already have dedicated short-range communication (DSRC) radios, and the city has plans to equip the remainder later this year. The 150 additional intersections will also have DSRC radios.
Right now, the company is running a simulation for another technology that would allow a vehicle to share its intended route with the smart intersections. A navigation device in the car could relay information to the intersection, allowing a car through the network up to 25 percent faster, Smith said.
“The more people that are equipped, the more people can get that benefit. The overall performance of the network just improves,” he said. “It sounds a little magical, but it’s actually really simple. It’s just that we’re giving the signal system more information, so we can do a better job of optimizing it. It doesn’t have to guess what vehicles are going to turn here, turn there.”
Work on many of these projects has become possible in part because of Smith’s results from the initial smart traffic lights.
“I think having the really strong results makes a big difference,” he said. “People are much more receptive to listening to you.”