Published on the 24/09/2024 | Written by Heather Wright

But it’s not always been a smooth road…

Transurban says it has slashed energy consumption, increased asset life, reduced costs and improved operational efficiency using advanced analytics and AI.

The Australian builder and operator of toll roads outlined a range of initiatives it has already deployed at Gartner’s recent IT Symposium on the Gold Coast, with head of data and AI, Artak Amirbekyan telling attendees AI has had a big impact on the company and is bringing significant benefits.

“Have you seen anything in AI that is future-proofed?”

Among the initiatives is a ventilation optimisation program, which saw a digital twin of tunnels created and advanced modelling and simulations run to optimise ventilation systems in the company’s tunnels.

Ventilation systems consume 70 percent of energy within the tunnels on Transurban’s road network.

Tanya Trott, Transurban chief technology officer says the company knew it had ‘a really big opportunity’ to fine tune the system and reduce emissions and energy use.

LIDAR technology was used to map the accurate geometry of the tunnels and ventilation systems, then used to create the digital twin, which was augmented with real-life traffic data. An airflow simulation model identifies optimum run times for the systems.

“Like a lot of organisations we had a net zero by 2050 target, and prior to that we also had a 10 percent energy reduction we wanted to achieve by 2023. This solution helped us achieve that and is helping us meet the zero emissions by 2050 targets,” Trott says.

It’s also helped regulate ‘more appropriately’ the in-tunnel air quality, and resulted in an 18 percent power saving for the company overall.

But there was also another, unexpected, benefit, Trott says. Because the expensive industrial fans are running much less, the asset life has been extended.

Trott won’t disclose financial benefits, saying that’s ‘sensitive’. “But we did get a real good commercial benefit.”

Meanwhile an AI enhanced automated license plate recognition system is helping ensure it accurately bills road users whose e-tags aren’t working. The company relies on number plate reading to identify users if the e-tags are not working for some reason, or if a user doesn’t have the tag.

“We have about 2.5 million trips per day, so if even a small percentage can’t be read, it is quite a big impact for the business,” Amirbekyan says.

Off-the-shelf license plate recognition systems had limitations, so Transurban developed an AI-enhanced system that detects and corrects potential misreads by combining multiple models and fusing information from various internal and external data sources.

Amirbekyan says the offering has reduced the number of images going to manual processing by up to 40 percent.

Wary of the rapid changes in technology – “Have you seen anything in AI that is future-proofed?” he responded when asked about ensuring use cases are future-proofed – the company looks to modular architecture so if something else comes up it can be plugged in.

The automated license plate recognition system is an API, with data sent to AWS.

“In five years’ time we can change that API. Just chuck it out and bring something else in [if needed].”

The company has had a data and AI practice for almost five years, though it previously ran under the advanced analytics moniker.

Trott says when the practice was first stood up, it didn’t initially get much traction.

Work in partnership with a ‘really good’ ally who ran the road asset in Queensland would show the value of advanced analytics and the financial returns and benefits it could bring to Transurban.

Today the organisation has an Enterprise Data Council which includes all general managers and a number of the ‘heads of’ who are responsible for various business units. It meets regularly to review potential projects.

Trott says the council is key in helping with the change management and people side of technology.

“Those [Enterprise Data Council members] are not technologists as a rule and it’s important for them to understand about advanced analytics, AI and data,” she says.

Before a project can even get to the council there has to be a Lean Canvas – a high level business case or plan on a page – which is validated by the Data and AI team as technically viable.

It’s the Enterprise Data Council who signs off and prioritises projects.

“What really shifted the dial for us was engaging the finance team right from the beginning,” Amirbekyan says.

“Everyone wants to do something with AI now, but if it doesn’t make sense there’s no point spending time there.

“It always starts with use cases.”

He cautions companies starting out with AI to pick the first use case carefully because if it goes wrong, it creates the perception the technology doesn’t work or is not good for the business.

“Pick something that can help deliver on the company strategy and what it wants to achieve and very much focus on value. Once you get that first hurdle done, it will be much easier because you’re going to create trust in the company – trust that the team know what they’re doing, can deliver and are valuable to the company, working with strategic focus and delivering value.”

He admits though that sometimes you do want to take a risk and invest in less certain use cases.

“That’s where experience comes in.”

Transurban has also put a focus on using data analytics to to improve safety on its roads.

“We’ve built this system – it’s like layers of cake, bringing more data in,” he says. The system pulls data including in-car telemetrics from users, road and traffic data – the company has ‘literally thousands’ of sensors at the roadside –  and weather information to create context for any incidents that happen and how particular spots can be made safer.

Automated incident detection, meanwhile, is enabling Transurban to detect and respond to more than 1,000 incidents a week within around six minutes.

The company has a team of 80 traffic control room operators monitoring 6,000 CCTV cameras across the network. They’re now augmented by the automated incident detection – essentially an algorithm that scans all the visuals coming out of the cameras and looks for changes in pixel lighting levels. If no change is detected for 10 seconds an alert is sent to a control room operator to investigate.

“There might be a broken down vehicle we need to send a response vehicle out to get the person moving safely again, or it might be debris on the road that needs to be removed or changed weather conditions, such as heavy rain or fog, limiting visibility and therefore we need to slow down the speed in tunnels,” Trott says.

“Deploying this technology is one of the reasons Monash Accident Research Centre has named our roads twice as safe as like roads,” she adds. “We’ve had a huge benefit out of this solution.”

AI has also been deployed for dashcam telematics road maintenance, enabling the company to more quickly pinpoint road issues, including those unseen to the human eye, and for commercial vehicle classification used for long term modelling and strategic planning.

“In Transurban, AI is a reality,” Amirbekyan says. “It’s not theoretical. It is deployed and used every day and it is having quite a big impact on our company and bringing significant benefits.”

So does the company analyse the results of its projects carefully.

“Yes,” quips Amirbekyan, “we analyse our bank account when we see how much we pay for energy costs!”