Can the internet of things save us from traffic jams?
By Nicole Kobie,
Traffic is getting worse. It doesn’t just feel that way, the stats prove it: commuters in 2014 spent an average 66 more hours stuck in traffic than they did in 2013, according to navigation tech firm TomTom. So when internet of things technology is disrupting every part of our lives, when will traffic lights be rethought and rebuilt?
Well, the traffic light revolution is already underway. It is all part of the promise of connected and self-driving cars, which allow data about individual journeys, routes and vehicles to be centrally monitored, controlled and systematised.
Autonomous intersection management
Once computers are in full control of our cars, do we even need traffic lights at intersections? That’s the idea behind AIM – autonomous intersection management – at the artificial intelligence laboratory at the University of Texas at Austin. Rather than stop at red lights, self-driving cars would schedule a slot through an intersection in real-time, speeding up or slowing down to ensure they’re in the right place at the right time – and not smashing into another car.
In black and white text, that seems eminently sensible. But it won’t be for the fainthearted – at least not until passengers have learnt to entirely trust their automated pilots:
For the idea to work, it would require roads to be mostly full of autonomous cars, says project leader Professor Peter Stone – and then it wouldn’t seem so terrifying.
“When I show people that video, I tell people to not envision themselves with their foot hovering over the brake or with their white knuckles on the steering wheel, but rather they’re in the back seat with the windows dark, doing a crossword puzzle or reading the newspaper, talking to family or whatever,” he says. “Once the driving is not a human task and people grow to trust the software controllers, people will also get used to the idea of cars going through the intersections.”
That said, he stressed that driving is now and will likely always remain a risk and reward equation, but he predicted that with AIM, “the efficiency gains will be so high that they’ll offset the perceived risk”.
How much faster will careening through intersections be compared to carefully stopping? The researchers compared AIM to heavy traffic on a major road, saying it would reduce delay by as much as 100 times – though that’s only at intersections, not total driving time.
It not only promises to remove time waiting at lights, but will cut fuel usage and emissions as well. “A lot of emissions and fuel usage are caused by acceleration, and our system will allow the cars to make a much more constant speed,” he says, adding that the savings could be “quite significant”.
But even with a complicated system such as AIM, intersections will still be safer than they are today, Prof Stone says, pointing out that a third of all fatal accidents happen at intersections.
“Intersections are already quite dangerous. When a computer’s doing the driving, even with all the cars going through without stopping, it’s going to be a lot safer than it is today.”
‘Are we there yet?’
Computers aren’t driving our cars yet and won’t be for some time, but there are some connected car projects that already claim to be easing the flow of traffic.
TomTom collects swaths of traffic data from its satnav devices but also used anonymised data from third party navigation apps, including smartphone maps. “We have agreements with a number of smartphone manufacturers, so they provide us with real time GPS feeds wherever their smartphones are,” says Nick Cohn, senior traffic expert at TomTom.
It also gathers data from telematics units installed in fleet vehicles as well as in-dash systems, giving TomTom a comprehensive overview of traffic flows. The resulting information on near real-time congestion is shared with customers, which includes road authorities who use it to plan traffic management as well as consumers.
“Most have camera data that doesn’t cover the whole network, so they use our data to supplement that and for deciding whether they need to switch to a different traffic signal scheme,” Cohn says.
When a driver hits a patch of congestion – a red zone of a smartphone or satnav map – it may be because of data that was collected, aggregated and distributed from connected cars in weeks or months past. Before ubiquitous connectivity, Cohn said the travel times seen by TomTom were very different than that given by road authorities such as the AA. As data improves, the numbers are merging, suggesting travel advice has become more accurate.
As cars become more connected – whether it’s through satnav or simply the smartphones in our pockets – better data in means we get better data out on the road.
Andy Stanford-Clark, distinguished engineer in IBM’s global internet of things team, pointed out that we can now pull in all sorts of data: not only GPS from cars and timings from traffic lights, but also air quality sensor data and images from cameras.
“On its own, [each] is of low value, but when merged together in the internet of things’ cloud processing platform, [we] can make sense of them and make actionable insights,” he says. “It might be to turn some traffic lights green quicker or send a text message to a car, or alert satnavs in the car to quietly change the routing so they’re now going somewhere else.”
It may seem like a small change, but consider the shift that’s happened in digital signs. Highways England used to simply warn there were “queues ahead”, but now tells drivers it will be 17 minutes to their junction, points out Giles Perkins, business development director for intelligent transport at Mouchel, which runs the National Traffic Information Service . “More data in and more data out can only be a good thing.”
Though traffic data makes it possible to see the movement of traffic in real time, and traffic lights themselves are operated algorithmically, it is still not possible to engineer a way of turning the lights green as you pull up.
“It’s easy to change the traffic lights,” ssays IBM’s Standford-Clark. “But ... you get this terrible interconnection of unintended consequences.” Your main route into a city may be clear, but every road feeding into it would be gridlocked. “It’s not a trivial thing to do.”
That’s why most light sequences are set via a longer term algorithm, taking into account other parts of the road network.
As we shift to more autonomous cars, that may have to change. One way self-driving vehicles may be introduced is “platooning”, with a lead car in control of a train or group of followers, handy for giving lorry drivers a break.
Signals couldn’t be allowed to change midway through a platoon, or it would leave stragglers behind. “Platoons would need to transmit their status to the intersection and the signal change would need to be advanced or delayed to treat the platoon as a single long vehicle,” said Alan Stevens, chief scientist and research director at transport firm TRL.
“Either the platoon could signal that it’s cleared the intersection or there would need to be infrastructure sensors to check the whole platoon is through.”
This is already happening in in a limited sense, noted Stevens. “At a local signal level, we can implement priority measures for ambulances, buses, etc - that’s a standard feature in some software and has been for years,” says Stevens. “However, giving priority to one vehicle makes things slightly worse for all others. So, there’s little point in giving one or two connected private passenger vehicles special priority.”
There are reasons to give some cars priority, and that’s being trialled by Newcastle. There, traffic lights are “talking” to motorists, sending messages to a device in car about obstacles or delays ahead, as well as helping them adjust their speeds to hit lights when they’re green.
“The system might advise a driver that if they travel at 24 miles an hour they will hit the next four sets of traffic lights on green,” says Newcastle University professor of intelligent transport systems Phil Blyte, announcing the project. “In more congested areas or particularly busy times of the day, then vehicles on key roads might be given priority in order to keep the traffic flowing.”
The system also gives priority to non-emergency vehicles, such as those transporting people between hospitals, cutting NHS fuel costs and improving patient care. So far, 20 traffic lights are using the system in Newcastle city centre.
Zombies ahead – and cyclists
As with any tech innovation, one of the biggest challenges is security. The best example so far is surely hacked construction signs in the US, with attackers warning of zombies ahead, but it’s easy to imagine how taking out traffic networks could shut down a city or otherwise wreak havoc.
“As more technology and software migrates to the cloud and is configurable over-the-air then the number of ‘attack surfaces’ (to use the jargon) increase,” saysStevens. “Simple hacking to clone a bus asking for priority is one level of threat – relatively easy but the implications are not too serious. Affecting whole intersection or networked intersections would be a much more major threat, but the designers are aware of this and try to take steps to avoid such problems.”
There’s another challenge: roads aren’t only used by cars. What about the bicycles, scooters and pedestrians hoping to cross the street? They have smartphones, so there’s “potential to do something with that,” said TomTom’s Cohn.
“I think it’d be great as a pedestrian and a cyclist if I didn’t have to push a button to cross and I didn’t have to wait, that I could also be optimised in my walking and my cycling,” he said. “Except that I tend to not navigate, so I’m not really telling anybody what’s my trajectory or route when I’m walking or cycling.”
That raises a question: what’s the goal of IoT-connected traffic lights? “I don’t think the main reason is going to be racing people through intersections, it’s going to be safety applications,” said Cohn. “Though it’ll also have a secondary effect of optimising for time, it’ll be really safety focused.”
There’s more than just those two options, of course: are we trying to cut traffic delays, reduce emissions, boost public transport or improve safety, or is there some other goal we haven’t thought of yet? “There’s a whole bunch of optimisation goals which may not be the ones you first thing of,” says IBM’s Stanford-Clark.
The best way to achieve it may not be via smart, reactive traffic lights, but by sending messages to drivers in cars, giving them useful information to react to and nudging them into better routes.
“It’s probably easier to change driver behaviour than it is to change the traffic lights.” Until computers take over driving, at least.