How Virtual Traffic Lights Could Cut Down on Congestion - CTTEC - Carnegie Mellon University

Tuesday, February 5, 2013

How Virtual Traffic Lights Could Cut Down on Congestion

Every new glimpse at the future of urban transport seems to be missing something pretty big. A car without, you know, human drivers. A trains of cars commuting on the highway with human drivers who aren't, you know, actually driving. The next peek ahead, according to computer scientist Ozan Tonguz of Carnegie Mellon University, gets rid of all the traffic lights.

At least the physical ones. Tonguz and colleagues are designing a road-efficiency system, based on emerging-vehicle-to-vehicle technology, called Virtual Traffic Lights. The idea is to shift traffic control from fixed street signals to the moving cars themselves. The result, says Tonguz, is an optimized traffic flow that should greatly reduce city congestion.

"To do this means that we can improve the life quality of New Yorkers, people in L.A., Boston, Atlanta, and other major cities tremendously," he says. "If you really want to take major stride in solving traffic congestion problems, for the life of me I don't know how else traffic congestion can be mitigated."

The basic world of Virtual Traffic Lights operates like this: as you approach an intersection, your car transmits data, such as location and speed, to other nearby cars. The virtual system processes this information for all the cars in the area, with the help of a lead car that changes every cycle, and determines your individual traffic signal. Instead of seeing a red or green light hanging in the intersection, you see it on your windshield and stop or go accordingly.

The first advantage to Virtual Traffic Lights is that every intersection with a car now automatically has a traffic light. That may not seem like much, but fewer intersections are equipped with signals than many people realize. In New York City, for instance, only about 24 percent of intersections have a four-way signal. As traffic lights become ubiquitous, road safety should dramatically improve.

The second benefit is a much better traffic flow. The algorithm that governs the virtual system can be written for total efficiency. If the system recognizes that no cars are coming from another direction, it can extend a green signal indefinitely. Likewise, at heavy intersections, it can give preference to the longest line of cars. Using similar technology to Google's driverless car, the system can also recognize the presence of pedestrians and bicyclists, and orchestrate traffic to suit their needs.

"This technology essentially buys you universal traffic control at every intersection," says Tonguz. "It makes traffic lights smart in the sense that it takes into consideration how many vehicles are there in each approach and in that way adjusts the cycle duration."

In recent simulations the system has performed quite well. The flow of cars in Porto — the second-largest city in Portugal, with 16 percent signal coverage at intersections — improved by 60 percent during a rush-hour scenario of one test [PDF]. The Virtual Traffic Lights system has even beaten congestion pricing models "hands down," Tonguz says.

Tonguz drew inspiration for the virtual system from nature. His idea was to capture the basic rules of self-organization practiced by biological species and engineer cars to cooperate in a similar manner. If a "colony of vehicles" can act more like a colony of ants, he says, "the lump sum behavior will be much better than we have right now."

The basic technology for Virtual Traffic Lights is already here. Car-to-car conversations can operate over Dedicated Short Range Communication at 5.9 Gigahertz — a radio system being tested and refined by the federal government. Tonguz expects D.S.R.C. to become mandatory for new cars soon, and he's working on a prototype to retrofit older models. (His work is being sponsored by the Department of Transportation, and Tonguz says he's also received funding from General Motors throughout his career.)

The biggest obstacle, says Tonguz, is getting the government to test the system in a real-world setting.

"In my mind the argument here is so compelling, the benefits so compelling, it's not a matter of if, it's a matter of when," he says.

Article courtesy of The  Atlantic