Green Driver

This is very interesting - from the Babbage column in the Economist - Green Light Districts:

MUCH of driving involves stress and attention—or inattention. Drivers are required to observe everything happening on the road around them while increasingly choosing to distract themselves with phones (smart and dumb) and in-car entertainment systems. Matt Ginsberg, the boss of Green Driver, aims to remove a portion of that stress by pulling data out of cities' traffic-management grids to provide drivers with intelligent predictive information about traffic signals and related matters delivered via smartphones and automotive computer systems. The system will make its official debut today at the Consumer Electronics Show in Las Vegas.

Dr Ginsberg says Green Driver's currently free Enlighten app for Apple's iOS and Google's Android will count down the seconds while a driver is at a standstill waiting for a traffic light to change from red to green. The app chimes by default when five seconds remain. The app works with live data in Portland, Oregon, and his hometown of Eugene, Oregon, and is in beta testing across the state of Utah, in Las Vegas, a Dallas suburb, and three cities in California: Arcadia, Pasadena and San Jose. He notes that to avoid safety issues, the app provides no information while a driver is in motion nor alerts them of when a green is switching to yellow or red. "We don't want to have a distracted driver issue," he says, or people rushing to catch a light.

Green Driver accomplishes this seemingly simple task through a large amount of behind-the-scenes data crunching. American cities typically work with one of six traffic-signal management firms that design the hardware and software that manage traffic flow. Dr Ginsberg says, however, that the data these systems produce are a massive and continuous flow of lights' current statuses, rather than reports of changes between states and imminent changes. His firm has worked with each vendor to give them software code that produces a simplified stream that can be more easily interpreted.
Predicting when a light will turn green sometimes relies on identifying the particular timing program in place for a signal: some lights run on recurring intervals. Others are triggered by time of day, or sensors embedded beneath the pavement and pedestrian buttons (despite the widespread notion that such "walk" buttons do nothing at all, which is sometimes true). Green Driver looks at signal programming and historical data to determine whether it can, with confidence, tell a driver reliably when the light will change. If so, the Enlighten app triggers the countdown and then the chime. If enough cars run Green Driver's software through smartphones or internal computers with cellular links (such as GM's OnStar system), the software could anticipate cars about to trigger pavement actuators on adjacent streets which would allow further accuracy in predictions.

Green Driver's product, which requires no new hardware, joins a burgeoning automotive computing sector. GM has equipped 30 models of its cars with its cellular data OnStar system, representing some tens of millions of cars on the road, of which about 6m American owners have active subscriptions, which require a large one-time or a smaller recurring fee. The firm Automatic launched a product in late 2013 as a combination of a plug-in module that works with a standard data port on U.S. cars and a smartphone app. The module monitors a driver's behaviour for rapid acceleration and braking, and for high speeds on the motorway to provide a combination of audio feedback and retrospective reports via the app that aim to retrain skills so as to use petrol more efficiently.

Enlighten is just the start of Green Driver's efforts, says Dr Ginsberg. With this data, he says the company can address three separate goals, of which just one is reducing the stress of sitting at a light with unknowable period of time of alertness (or a penalty of drivers behind one honking or even rear-ending one). He notes that with integration into a car's computer, a hybrid could more efficiently choose between petrol and battery as it approaches an intersection where a light is about to signal a stop. He estimates a 3-5% saving in petrol from this one change, based on testing. The firm is talking to carmakers already.

There's also a safety angle. Dr Ginsberg suggests the case in which a driver is about to run a red light, either traveling too fast to catch the change or breezing through an intersection. If his software were continuously talking to the car, the auto could honk its horn and flash its lights (rather than, say, slam on the brakes), warning other drivers and potentially reducing side-swipe accidents. Data could also be anonymously collected and uploaded to tell local law-enforcement about trouble spots where drivers are blowing through lights, whether accidentally or intentionally. Green Driver can also provide fault data to cities about whether or not their signal-control systems are functioning correctly or optimally.

Dr Ginsberg says he is frankly baffled that his tiny company is the first and so far only to build such processing and delivery software. He must approach cities one at a time to make deals, and he thinks the foot-leather portion may have deterred potential competitors. His other firm, On Time Systems, has spent 15 years providing optimisation software for large-scale problems often seen as intractably irreducible, such as shipyard scheduling and efficient multi-point routing of aircraft, and is providing the expertise under license to this new effort.

Green Driver's work stands in contrast to one promoted by the United States Department of Transportation (USDOT), Dedicated Short Range Communications (DSRC) Technology, in which the USDOT proposes a new wireless networking system that would aid with traffic flow and attempt to reduce collisions, but also require new hardware to be built into every car and intersection, and new spectrum allocated.

The programme's work began in 2008, at a time at which American cellular data networks still had relatively low capacity, and only several million true smartphone users. Between the 145m Americans who own smartphones (61% of mobile phone owners), GM's OnStar, other carmakers' products and self-driving cars, the red light may stay on for the USDOT effort, while fleeter firms like Green Driver get the signal to go."