Published Aug 18, 2020
A Victoria University of Wellington PhD student and his supervisors are developing a game-changing machine learning protocol that could help to predict digital network failures before they actually happen—significantly reducing network downtime as a result.
It’s also the first project to be taken under the wing of Wellington UniVentures’ incubator programme, to help the researchers bring their idea to market, where it can achieve the most impact.
“We’re excited about the support and capability the incubator can offer the University’s entrepreneurs,” says Dr Ashwath Sundaresan, the Wellington UniVentures Senior Commercialisation Manager who is also responsible for overseeing the incubator programme. “We’re able to intercept ideas and provide inventors with wrap-around support much earlier, which helps to get their research investor-ready faster.”
He says that because product development and startup management can be a little ‘out of left field’ for most researchers, the incubator has been designed to ease the transition from early-stage research into functioning startup.
“We offer the University’s inventors a real depth of capability that includes help with a full range of business skills including marketing and product development etc.”
That level of help is expected to expedite the process of turning the idea into a product that could ultimately change the way network operators monitor and maintain their networks.
“Network downtime causes all sorts of costly issues for the owners of the network,” says Dr Paul Geraghty, the Wellington UniVentures Commercialisation Manager responsible for this project. “However, detecting network issues can be like looking for a needle in a haystack— particularly for regional network providers whose physical assets may be widely spread.”
He says that finding and fixing a network problem is still a time-consuming, manual process that requires people to hunt through mountains of data to identify the source of an anomaly—by which time it may have already caused a critical outage.
PhD student Murugaraj (Muru) Odiathevar and his supervisors—Professors Winston Seah and Marcus Frean from the University’s School of Engineering and Computer Science—are developing an automated solution: a unique Hybrid Machine Learning (HML) methodology that constantly learns about new data streams and anomalies in digital networks—enabling network issues to be predicted before they even happen.
Muru says that while networks can already be trained to detect known issues, the constant flow of new data means that networks struggle to determine new issues as they arise.
“What does ‘normal’ look like when ‘normal’ is constantly changing?” says Muru. “What we’re doing is essentially teaching machines to adapt and respond over time to evolving network issues in a way that’s never been done before. This will become increasingly important as we see more devices connect to networks.”
Singapore-born Muru says that while he has just joined the incubator programme, he’s already seeing the value. “I only came to New Zealand two years ago, so I didn’t have a lot of contacts here,” he says. “But thanks to the incubator, I’m being introduced to the people and organisations that will be a great help to us in developing a product that fits industry needs.”
One of those organisations is KiwiNet, who not only provided funding to develop a prototype—which the team is trialling in a regional broadband provider’s network—but also welcomed Muru and his fellow team member and PhD student, Duncan Cameron, into their Emerging Innovator Programme. This represents the first time a team has been invited into the programme, rather than an individual.
Muru says they are now looking to connect and partner with other digital network operators across a broad range of industries—for example, energy supply, rail or aviation—for a second round of trials to ensure they understand a broader range of industry needs before taking it to market.
“Giving network operators better visibility on potential failures will lead to reduced network downtime, and increased network resilience by preventing critical failures before they occur,” says Muru. “I’m grateful for the opportunities I’ve been given to develop this research further and potentially make a real difference in the world.”
For more information about this project, please contact Dr Paul Geraghty (below) or to learn more about the incubator programme, please contact Dr Ashwath Sundaresan (below).