Smartphone system to predict traffic light sequence

A new system being developed in the US aims to take the chaos theory out of traffic management — and it's all done with smartphones.

If you’ve ever seen overhead footage of congested city streets, you’ll have noticed a compression-release effect at every intersection. The lights go red, cars stop, more cars bank up behind them, the lights go green and they decompress into the empty space ahead left by the traffic that made it through the lights. Until the next set of lights, where the cycle repeats.

The movie Koyaanisqatsi sped up the footage to make art of it. But every time that happens, a whole bunch of vehicles is corralled into a stop-start manoeuvre – and it happens dozens of times a minute in any city CBD. Given internal combustion engines are at their dirtiest and thirstiest taking off from standstill (more so when they’re cold), that puts a serious dent in every city car’s fuel efficiency.

The best answer yet devised (besides taking a bicycle instead) is the Green Light Optimal Speed Advisory (GLOSA) system, which uses traffic signal data to let drivers know when to slow down in order not to have stop at an approaching set of lights. The idea is simple: if everyone slows down, it reduces the number of cars that have to stop. The result is a reduction in aggregate fuel consumption and emissions.

But GLOSA systems have not set the world on fire. There’s only a few in operation, thanks largely to their complexity and the costs of rollout and maintenance.

But change is in the airwaves. With interest in such systems on the rise, researchers from Massachusetts Institute of Technology (MIT) and Princeton have unveiled an advisory system based on input from drivers’ smartphones.

SignalGuru uses images from phone cameras to record lights turning red and green and predicts subsequent signal activity to advise drivers when to slow down to minimise stop-starts at lights and smooth traffic flow.

Testing by the team has shown that by reducing idle time and stop-start, SignalGuru can cut aggregate fuel consumption and emissions by around 20 per cent. The system uses a network of smartphone cameras mounted on car dashboards to collect information about traffic signals.

The SignalGuru team’s research concluded that the imposition of constant stop-starts not only annoys drivers and lengthens commute time, but increases fuel consumption by 17 per cent and CO2 emissions by 15 per cent.

GLOSA systems feed off a constant flow of traffic signal data. SignalGuru’s big advantage lies in its ability to operate independently of traffic management infrastructure – it works entirely on the local cellphone network.

In presenting its award winning paper at the Association for Computing Machinery’s MobiSys conference, the team described how two real world deployments – in Cambridge, Massachusetts and Singapore – showed up both its considerable benefits and its equally considerable shortcomings. It managed to predict the schedule both for old-school pre-timed signals and those using current adaptive technologies.

Most importantly, the study’s data collection extended down to measuring fuel efficiency benefits. A test vehicle working by the system turned up fuel consumption reductions of 20.3 per cent.

The team admitted the system has its share of shortcomings. GLOSA systems will always work best with input from the network of loop detectors feeding into public traffic management systems from under the road surface at lights. The smartphone cameras comprising SignalGuru’s critical data gathering tool are normally of pretty low quality, and they’re feeding in through a channel with limited computing power and bandwidth.

Thanks in part to these constraints, the system’s signal detection algorithm has also proved mistake- prone in an environment with little margin for error. By the team’s reckoning, a misdetection rate as low as 4.5 per cent is all it takes to corrupt up to 100 per cent of traffic signal predictions.

Many of these issues will be resolved as phones gain more processing power and network bandwidth grows. And in time, authorities will start incorporating GLOSA systems into public infrastructure. Agencies in the US and Europe are already lobbying to have short range transceiver antennas built into traffic signals in expectation of something they see as inevitable.

In the meantime, SignalGuru has shown up the advantages of GLOSAs. And besides, much of its value lies is in the system’s adaptability to other uses. The team mentioned using it to capture and disseminate all kinds of information useful for commuters, such as petrol prices at different stations, about where buses are on their routes and what kind of time they’re making, and about parking availability.

Read the latest news and reviews on your mobile, iPhone or PDA at carsales' mobile site...