Pittsburgh is slashing emissions and journey times by using real time traffic flow data to determine when traffic lights should go red or green. Each intersection builds and updates its own lighting plan every few seconds by using camera and radar data to calculate where vehicles are, when they will reach an intersection and how long it will take them to pass through it. Traffic is logged in time-sequenced clusters of vehicles, rather than as individual cars, allowing the system to accurately monitor flow, which is then passed on to adjacent intersections.
Results & Impact
Where the smart traffic light technology has been installed, intersection wait times have fallen by up to 40%, journey times by as much as 25%, and emissions by up to 20%.
Scalable Urban Traffic Control (SURTRAC), Carnegie Mellon University, Rapid Flow Technologies, City of Pittsburgh, Hellman Foundation, Richard K Heinz Foundation, University of Pittsburgh Medical Center
The system monitors the volume of traffic approaching each intersection in real time through cameras and radars. They can detect vehicles about 100 metres away and transmit their location to a computer installed at the intersection. The traffic location data is then logged in time clusters, allowing all vehicles within a certain distance of each other to be represented through a single data point on the computer system. Each stretch of road is split into zones, and the time it takes for vehicles to pass through each one is monitored. This allows the system to calculate variations in traffic flow throughout the approach to each intersection and predict how the traffic will flow. This aggregated picture of traffic is used to calculate light sequencing plans, while the expected outflow of traffic from each intersection is transmitted to the next traffic signal on the network.
Road users, city dwellers
Cost & Value
In 2016, Pittsburgh received more than $20 million from the federal government and state authorities to increase the number of SURTRAC installations from 50 to 200.
Running since 2012
Although the system can predict what the overall flow of traffic is likely to be, it cannot account for the actions of individual cars. For instance, cars may change direction or enter the road between intersections, and so would not be accounted for in anticipated flow rates.
Atlanta is set to implement the SURTRAC system at 25 intersections in 2017.
Pittsburgh has used smart traffic lights to cut car emissions by 20% and travel times by 25% in five years.
The Scalable Urban Traffic Control (SURTRAC) system, implemented in Pittsburgh in 2012, uses real time traffic volume to determine whether lights should go green or red. It has also cut the waiting time at intersections by 40%.
In SURTRAC, sequencing plans are developed by traffic lights at each intersection. This is done by tracking the locations of vehicles through cameras and radars, allowing computers to predict how long it will take them to reach the junction and when they will pass through it. Traffic is logged as they leave an intersection and the data is shared with adjacent lights, allowing each traffic light to build up a picture of vehicle flow long before it arrives.
“The key to coordination is that we can build long-horizon plans,” said Professor Stephen Smith, Director of the Intelligent Coordination and Logistics Laboratory at Carnegie Mellon University, which developed the SURTRAC system. “This formulation allows us to solve the intersection control problem orders of magnitude faster than what was previously possible.”
In conventional signalling systems, lights operate according to set timing plans designed for the average volume of traffic expected in a given period. For instance, cities may operate different sequencing plans for the morning, afternoon and evening. However, variations in traffic flow cannot be incorporated into these models.
“These plans start to age as soon as you install them,” said Smith. “Traffic patterns evolve over time, particularly where you have development.”
Rather than having a standardized plan covering all lights in a particular area, SURTRAC allows each light to build its own plan. Anticipated traffic flows are then shared with adjacent traffic lights either by fibre optic cable or point-to-point radios. As a result, each installation can create a light sequence encompassing overall traffic volumes within the network and local variations within its particular section.
“We take a totally decentralized approach,” said Smith. “Each intersection controls its own local situation and then intersections interact with their neighbours to coordinate activity through the network.”
Key to the SURTRAC system is its use of clusters to aggregate the positions of vehicles. Rather than mapping the position of each vehicle individually, their locations are recorded in time clusters on computers installed at each intersection. This means that vehicles within a certain distance of each other – for instance, one second apart – are aggregated and represented as a single group. This is combined with a zoning system, whereby each stretch of road is broken down into blocks, allowing the position of cars to be recorded in relation to different parts of the road, as well as other clusters. This allows it to predict how large vehicle clusters will become, how long it will take them to reach an intersection, and when they are likely to clear it.
“It’s really that aggregate representation, that’s what allows us to build these long horizon plans very quickly,” said Smith. “What’s key about it is it preserves the nonlinear pattern of traffic. So you still get to see the spacing of vehicles but it’s just on an aggregate level. If you’ve got two clusters and a fair amount of time between them, that’s an opportunity to switch phases and process clusters north-south as opposed to east-west.”
The system was developed by researchers from Carnegie Mellon University (CMU), who have since commercialised the technology through a spin-out company called Rapid Flow Technologies. To build and test the system, CMU received timing plans and traffic volume data covering several intersections from the city of Pittsburgh. The CMU team then built two models to run different simulations of the same data, one showing traffic flow with the city’s existing system and the other showing how flow rates could be improved through CMU’s aggregated system. The model’s impressive results persuaded the Hellman Foundation, a local philanthropic organisation, to fund a live trial in 2012, which confirmed the potential benefits.
Since then, Pittsburgh has steadily expanded the number of intersections using the SURTRAC system. Beginning with nine intersections in 2012, the city doubled the number of installations in 2013, and has since expanded it to 50.
The system could soon become even more accurate thanks to technology that would allow cars to communicate directly with the SURTRAC computer systems. Dedicated Short-Range Communication Radios (DSCRs), increasingly installed in cars, transmit vehicles’ location, heading and speed data every 10th of a second. SURTRAC can process this information and notify the vehicle when a light is about to change colour, allowing it to adjust its speed to align with any change in lights. Pre-planned navigation routes can also be shared with SURTRAC via DSCRs, telling the system the exact route a vehicle is planning to take through the network. This can improve journey times by a further 25%.
In 2016, Pittsburgh received over $20 million from federal and state authorities to triple its network, with which it hopes to install a further 150 SURTRAC lights by 2019 or 2020. Atlanta is also planning on launching its own SURTRAC system, with a network composed of 25 installations.
(Picture credit: Pixabay/doctor-a)