Data Analytics to Help Clear Auckland Congestion

Traffic congestion and underperforming transport infrastructure are not only annoying for commuters, but it also costs the economy money through inefficiency.

In 2017, a group of NZ business organizations got together and commissioned a report from the New Zealand Institute of Economic Research which showed that Auckland would get an economic boost of between NZD 900 million and NZD 1.4 billion if it had a free-flowing road system.

According to GPS navigation company TomTom, Auckland ranks 40th in the world in terms of traffic congestion, and the average resident wastes an average of 45 minutes each day in their car, worse than a much bigger city such as London.

Right now, immediate improvements are not easily achieved. As the city grows, infrastructure projects such as a new light rail project, busway extensions and a new motorway are being added, but they will take time.

Analytics to the Rescue

To fill that gap, Chris Creighton said that Auckland Transport, the local government authority charged with operating trains, buses, ferries and the city’s arterial road system, is turning to data analytics to make the existing infrastructure function more efficiently.

“Many of these improvements will take years to happen, and right now is the time we can fill in some of that gap with data projects that can make a difference,” said Creighton, the Group Manager of Business Technology Solutions for Auckland Transport at a Gartner Data Analytics event in Sydney last week.

“We are finding that even small investment can give you a marked increase and a better performance around our key criteria which are customer experience, mobility and safety.”

Creighton outlined three digital case studies undertaken by Auckland Transport which had enjoyed some success: a machine learning project on predicting delays and managing transport connections, a video analytics project aggregating data from 3,500 CCTV cameras around the city, and a “dynamic lane” trial on a major road using changing LED lights.

The organization has also concentrated its data team in an analyst hub, combining data scientists and analysts working on multiple projects.

Keeping Things in Motion

The first project is called “things that move” and harnesses machine learning. It came from the understanding that 40 percent of public transport users take multiple modes of transport in Auckland.

The system is configured as a “hub and spoke” with bus services feeding into larger train and ferry services to major destinations, such as the Auckland CBD.

“When the connection doesn’t work, that is not a good outcome for our customers,” said Creighton.

Using a Microsoft machine learning solution, Auckland Transport has aggregated historical data on passenger numbers and the impact variables such as holidays and weather can impact on the timetable, and on the ability of people to make their planned connections.

Using the example of a ferry leaving the Auckland CBD, controllers have an accurate understanding of the number of people using that ferry who will need a bus for further transport to get home, and they will also understand the number of people waiting at bus stops along the route at that particular time of the day. The system can predict this with an accuracy of 80 percent.

"We have built up that model, and it gives predictions to our control staff, who then have the ability to hold services so that the connections can work in real time,” said Creighton.

“We are also publicizing these decisions to our customers, so if we do decide to hold the service then we send that information out over the various channels we have, such as our web, our passenger information services and our apps, so that we get that information about delays and rescheduling out there,” he continued.

“Our customers have told us that the perception of a delay is worse. If you tell them there is a delay that is ok, but if you don’t then a five-minute delay feels like ten minutes, and ten feels like twenty,” he added.

While the results are so far anecdotal, they have been positive in terms of customer feedback.

“If they have trust in the information you are providing, and the bus does get there then it builds confidence,” said Creighton.

“So instead of customers missing connections and relying on paper-based information we are integrating transport modes more efficiently and notifying customers in real time.”

Improving Decisions, Easing Congestion

In the second project, Auckland Transport is applying analytics to the vision fed in from 3,500 CCTV cameras.

When something abnormal occurs, such as a car stopped on the road at a crossing or people walking through an intersection while alarms are on, this is brought to the attention of monitors who can then take action.

If there are high numbers of people in cars running red lights at a particular intersection, then further action can be taken to understand the cause, such as the phasing of the lights or the physical design of the crossing. 

"This analytics is adding to the value of what we can do," said Creighton.

“For the future, we think we can use this technology to optimize the sequencing of traffic lights and improve traffic flow for the whole city.”

The third was a pilot project creating “dynamic lanes” using LED lighting.

Lights embedded into the road surface marked lanes, which could be changed quickly and safely to create temporary lanes during times of heavy congestion, with the information reinforced on traffic control gantries overhead.

The trial ran for 12 months until January 2019 and has been labeled a success, and the concept is likely to be rolled out on other roads through the city.