Whenever Lizzie Ward commutes into Cambridge via one of the main arterial roads off the M11 that leads directly into the city centre, she finds herself frustrated.
“Lights have recently been installed at the turn-off for Eddington, a new housing suburb created for the university,” says Ward, an occupational therapist. “The problem is this set of lights is only yards from another set and they don’t appear to be sequenced. This is causing long tailbacks to build up at peak times. It would make more sense if the lights were synchronised and the length of time they stay green changed according to the volume of vehicles on the road.”
Ward’s sentiment is no doubt shared by many who commute into Cambridge. The city was found to be the UK’s gridlock capital in a study carried out by telematics products provider Satrak. Over the course of 2016, data was pulled together from tracking software installed in 527,000 cars, lorries and vans, to determine how quickly or slowly traffic was moving.
Cities, including Cambridge, are realising they need to get smarter when it comes to traffic signalling systems and are beginning to invest in better technology. With the help of machine learning and artificial intelligence, traffic lights could eventually get better at predicting driving patterns and behaviours, reducing delays, accidents and emissions. And the ultimate aim is to communicate road advice to drivers.
A smart traffic lights trial is currently underway in Manchester (Credit: iStock)
Joining junctions
First, though, traffic lights need to be automated and connected to a traffic management centre. Currently, only 40 of Cambridge’s lights, less than a quarter, are automated. They rely on sensors under the road, known as induction loops, to control traffic flow and ease congestion, says the Greater Cambridge Partnership’s interim transport director Chris Tunstall.
Induction loops use wire coiled to form a loop under the surface of the road. The loop resonates at a constant frequency when no vehicle is present, but the frequency increases when a car passes over. The closer its metal undercarriage is to the ground, the greater the frequency increase will be. That data is sent to the traffic light control system so it knows when there’s oncoming traffic or a vehicle waiting at the lights.
A central computer uses the data to vary signal timings in real time to improve the progression of traffic. If, for instance, an induction loop is installed further down the road from a roundabout or junction, it’s able to detect if queues are building up and the system can hold the green light for longer.
As part of a potential £500m investment in transport in the city, the Greater Cambridge Partnership has been exploring upgrading and networking its traffic lights into a Scoot system. Scoot, an acronym for Split Cycle and Offset Optimisation Technique, is one of the world’s leading adaptive traffic management systems. It coordinates the operation of lights in a given area – mainly those at junctions and roundabouts – and responds according to how traffic flow is fluctuating.
“This would usually require each light to be connected to an induction loop,” says Edward Leigh, who leads the Smarter Cambridge Transport project, a non-commercial initiative pushing for better urban infrastructure.
Scoot is a more effective approach than relying on lights that have a fixed timer, says Leigh, because, even though the latter can be optimised for different times of day and recurring traffic conditions, planning is required. And because the lights don’t automatically react to changing traffic patterns, they have to be updated regularly – which can be expensive.
Introducing Scoot across the city would allow Cambridge to adopt a traffic management strategy known as ‘inbound flow control,’ also known informally as ‘queue relocation’. “Lights on the edge of the city would be able to release vehicles only as fast as the road ahead can carry them,” Leigh explains. “By restricting flow at peak times in this way, congestion would only be experienced, if at all, at the edges of the city.” Instead of traffic sitting in queues and blocking junctions within a city, it would be relocated to the edges where there’s plenty of space to hold it.
While Scoot is considered a smart system, it was originally developed in the 1980s by the UK’s Transport Research Laboratory. To date, it is integrated, either fully or partly, in more than 200 towns and cities worldwide – Seattle rolled out its version of the system in April. But, as effective as the Scoot system has been, some researchers and innovators believe that lights should be able to think for themselves.
The V2If device monitors chatter between smart traffic lights and control centres (Credit: iStock)
Lights that think
About 45 miles away from Cambridge in Milton Keynes, mobility start-up firm Vivacity Labs is piloting a project where 2,500 cameras, with artificial intelligence built in, will be installed in traffic lights. The project, which has received more than £3m in funding, will complement the city’s MK:Smart initiative, responsible for the UK’s first city-wide Internet of Things network and a hub for autonomous transport systems and intelligent mobility projects. The project is in its initial data gathering phase, which will last until mid-to-late 2018, with integration into traffic management systems to follow.
The problem with most lights, says Yang Lu, co-founder and chief technology officer of Vivacity Labs, is that, although they are at least sequenced, and some are automated to respond to realtime conditions, “they’re not truly smart”.
Lu believes that if lights were able to learn about driving behaviours and have eyes on all the roads at once, they could think ahead and make predictions. If there’s an accident, for instance, drivers could be alerted to an alternative route where green lights are being held for longer. The data collated by Vivacity’s sensors would be more refined and less crude than that gathered by Scoot, he adds.
A similar project, led by Rapid Flow Technologies, is under way in Pittsburgh, Pennsylvania, where 150 intersections will be fitted with smart lights over the course of the next three years. Each set of lights at an intersection will make its own decision, then relay information it gathers to the next intersection, and so on.
Cars that listen
Successful implementation and longevity of projects such as those taking place in Milton Keynes and Pittsburgh depend on two things: technology that relays information between vehicles and infrastructure, and cooperation from local authorities to share data.
That’s according to Matthew Ginsberg, founder and chief executive of Connected Signals, an Oregon-based company that has built a smartphone app, EnLighten. It predicts when lights will turn green by feeding data from connected traffic lights into a learning model that takes realtime conditions into account, and then sends that information back via cellular networks.
The main technology that’s being explored to enable vehicles to communicate with traffic infrastructure is called radio-based direct short-range communication (DSRC), but Ginsberg says it is outdated.
In Atlanta, Georgia smart traffic monitoring includes thermal cameras (Credit: iStock)
“Infrastructure has advanced since the DSRC approach was first envisioned a decade ago. Traffic lights are now typically connected to urban traffic management systems and we’re seeing an increasing number of vehicles connecting to the internet over cellular,” he says.
At the end of 2016, Audi announced that a selection of its 2017 models in the US would be fitted with DSRC. And Cadillac’s CTS now comes equipped with the technology. Despite this, some in the automotive and technology industries, including Ginsberg, believe that its deployment is not an economically viable solution and, given the emergence of 5G, is only a stop-gap measure at best.
“It costs about $20,000 to prepare an intersection of already connected lights for DSRC radio. This would total tens of millions, pounds or euros, in a city with a population of a million or more, which doesn’t sound like a wise investment to me,” says Ginsberg.
To counteract the problem and to support its EnLighten app, Connected Signals has developed a Raspberry Pi-based device running on the Linux operating system which, says Ginsberg, “leverages existing infrastructure to reduce cost”.
The device, known as V2If, can be used by local authorities to monitor communications between traffic lights and the traffic management centre via an Ethernet port. The communications are processed to determine future signal changes and the information is then sent to drivers over cellular networks.
“This device costs just $60 to install per city. We also bear the costs associated with maintenance and support. Otherwise it would be much more expensive if authorities had to purchase them,” says Ginsberg. He adds that one of the main reasons for the minimal costs is that Connected Signals is more interested in the value that can be aggregated from traffic light data. But to acquire the information, the company needs the support of municipal authorities.
Around two-thirds of towns and cities across the US aren’t connected to a traffic management centre and so for them the industry-preferred DSRC will remain the mechanism of choice for now. Ginsberg is confident that more urban areas will switch to smart traffic light systems in the future, giving Connected Signals an opportunity to access more data. With the reduced costs associated with the V2If device, it could prove to be an incentive for cities to switch to smarter signalling systems – connected to traffic management centres – if they haven’t already.
“I think, overall, current traffic light systems are pretty good. There are obviously improvements to be made in terms of them adapting to traffic conditions and better optimisation of signal timings. But these improvements are coming,” says Ginsberg. “The fact that it isn’t standard for vehicles to know what traffic lights are doing is an impediment to smarter driving behaviour, but, again, these changes are coming as well.
“So while lights aren’t talking and the vehicles aren’t listening right now, the arrival of connected and autonomous vehicles will help drive this forward.”
What happens in Vegas... is that traffic lights are "talking" to computers onboard selected cars (Credit: iStock)
Going green
Back in Cambridge, Leigh is hopeful that the introduction of Scoot will lessen transport pains. “Smoother traffic flows will reduce the likelihood of junctions and filters being blocked, which inevitably and quickly leads to gridlock and increases the risk of accidents,” he says. “And with vehicles having to make fewer stops and starts, there’ll be an improvement in air quality.”
In 2014, a team of researchers at Newcastle University highlighted how cities were underestimating emissions levels by up to 60% by monitoring the average speed of traffic as a whole, and assuming vehicles were travelling the same speed all the time. The researchers argued that taking stop-start emissions into account could help the UK reduce the number of premature deaths each year associated with traffic pollution. According to data published by the European Environment Agency last November, the UK had 11,940 premature deaths in 2013 from NO2, a toxic gas associated with diesel vehicles.
“It can be incredibly infuriating if you’re being held at a red light, then travel a matter of yards down the road only to be stopped at another red,” says Ward, whose journey to work could soon get smoother thanks to smart technology. “Anything that helps to reduce the need to stop-start will be welcomed by drivers.”
* Content published by Professional Engineering does not necessarily represent the views of the Institution of Mechanical Engineers.