Trinity College Dublin’s Marco Ruffini says digital replicas of networks can help analyse the outcome of a set of inputs and make predictions without affecting the actual network.
Automation has grown exponentially in recent years. In fact, a report from software company Workato found that automation adoption for processes such as insights and analytics in the EMEA region is up more than 400pc in the last year alone.
But while some automation technology can be easily integrated into a business, others require a certain level of testing before rolling out in the real world. This is particularly true in areas such as manufacturing and pharmaceuticals.
For these situations, the growth of digital twins will be an exciting evolution. A digital twin is a virtual representation of a physical product or process, used to understand and predict the physical counterpart’s performance characteristics.
Last year, Dublin City University announced that there would be a digital representation of the university’s campuses, incorporating real-time data on footfall, congestion points, energy, water usage and other important data points to inform planning and development of infrastructure.
And earlier this year, the European Commission launched a new initiative to develop a highly accurate digital model of Earth to help monitor, model and predict natural and human activity, and develop and test scenarios for more sustainable development.
Digital twins in telecommunications
Digital twins can also be used in the telecommunications sector to test changes to a network on a simulator before rolling them out on the real network. To learn more about this application, SiliconRepublic.com heard from Trinity College Dublin’s Marco Ruffini.
Ruffini is a principal investigator at both Connect, the Science Foundation Ireland (SFI) research centre for future networks and communications, and the Irish Photonic Integration Centre (IPIC).
He also leads the optical network and radio architecture group at Trinity and the new Open Ireland research infrastructure, which brings next-generation technologies like OpenRAN, OpenOptical, edge cloud and intelligent network control to a city-wide testbed.
‘How to build a reliable digital twin for a network is still an open question’
– MARCO RUFFINI
“We work on intelligent algorithms for controlling networks. An intelligent algorithm can monitor network performance, such as signal quality across different areas, and predict how this is likely to affect its users,” he said.
“Then this information can be used to react to changes in signal quality to maintain service quality. This applies whether the network is a mobile 5G or a fixed optical transmission network, although the type of service and user would be different.”
A digital twin of a network can be used to predict how the real network would react to certain situations, such as an accidental cable cut or the intentional addition of a transmission channel.
“A digital twin can analyse the outcome of a set of inputs and predict its outcomes without affecting the actual network. If the outcomes are those intended, then these new configurations can be transferred to the real network,” said Ruffini
While the idea of simulating network behaviour has been around for a long time, digital twins bring in the concept of a data-driven approach to run machine learning algorithms that can better approximate the complex behaviour of a network.
According to Ruffini, one of the major challenges is how to architect a digital twin, building it from standalone machine learning components. “There has been substantial work on the use of machine learning to model network elements or specific features, but how to build a reliable digital twin for a network is still an open question,” he said.
“Another challenge stems from the use of machine learning itself as a modelling tool. Many algorithms can approximate even complex network behaviour well, but they operate as black boxes, meaning that it is difficult to predict if there are conditions in which the algorithm might fail, thus providing non-optimal or wrong predictions.”
From virtual campuses to a digital replica of the Earth, it is easy to see how digital twins could be seen as game-changing technology. However, Ruffini said there is still much work to do.
“It is difficult to predict a time frame or what level of confidence will an actual digital twin deliver in practice, but the groundwork is already there, from investigation of machine learning models to the development of open interfaces towards the network component that enable a digital twin to directly interact with the network hardware.”
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.