How wireless, AI are driving a digital twin revolution

When someone says the word “wireless,” the first thing that likely comes to mind is probably a smartphone. But wireless technology also has permeated our homes, from smart appliances to personal voice assistants,  as well as nearly every device we use, from personal drones to automobiles.

“It’s the communication link that supports almost everything that we do,” said Walid Saad, professor of electrical and computer engineering at Virginia Tech in the greater Washington, D.C., metro area.  

Saad is innovating a wide range of applications for NextG technology and wireless cellular systems. Along with Lingjia Liu, director of Wireless@VirginiaTech, they are creating a wide range of applications for NextG technology and Open-Radio Access Networks.

Saad and his team are exploring, through work with international partners at the Institute of Science Tokyo, what wireless might be capable of next. Specifically, Saad believes that advancements in other technologies, like artificial intelligence (AI), can help drive more advancements in wireless and vice versa. And it all has something to do with digital twins and the digital world.

While we all lean on wireless technology every day, the generations that we hear referred to in shorthand — 3G, 4G, 5G, and beyond — the details of those technologies are a bit more arcane. When 5G was ushered in, it promised several things: a higher rate of data, a higher reliability of its transfer, and a decrease in latency, or delays. But, Saad said, to date, the reliability only really delivered in lab settings. That changed the goalposts. With an adversarial environment and space getting busy, given so many devices using the same channels, resilience became the key barometer. Rather than ensuring an ultra-reliable connection in all conditions, the focus shifted to enabling a quick recovery of network performance after a possible degradation.

For a phone call, this might mean the insignificant exchange of having to repeat a few words, rather than dropping the call entirely. But when applied to advanced technology, such as autonomous vehicles moving through a city or drones navigating treacherous terrain, a similar temporary lapse can have devastating effects.

The latter is what Saad and his team are working on with researchers from the Institute of Science Tokyo, formerly Tokyo Tech, through federal funding from the National Science Foundation. While this particular test bed is working on autonomous cars, the technology might be more widely applied in factories and for other industrial uses, where even marginal gains in efficiency can yield significant returns.  

That involves creating digital twins, or digital recreations of physical assets, that can be tested and monitored in parallel with their material counterparts.  

To fully understand digital twins requires clearing up a common misconception, or conflation: The digital world is not the same as the virtual world. Virtual reality can show a visual projection of what something looks like, whether through glasses or technologies like augmented reality. But the existence of those images is merely a projection or a simulation. It doesn’t live in any physical space.  

In contrast, a digital twin lives in a digital space parallel to the analog world. It can relay real-time control back to the real world. It also has to represent the physics of the real world so it can foresee plausible futures of the entire world. So, on a mass scale, digital twins would need to live in a fully realized digital landscape connected in all the same ways of their real-life counterparts.

Fueling an entire digital environment, even in a somewhat contained space like the college campus in Tokyo, requires massive computing power. Scaling up and connecting those environments together expands that need exponentially. Which is why Saad believes AI could actually be the key to wireless progress.

Advancements in computing and AI technology allow for much more data to be processed more quickly than ever before. But that data still has to be transmitted accurately and without delay from sender to receiver in order for it to be effective for technologies like autonomous vehicles. That, in turn, requires a local, digital replication of the real world environment, acting and reacting in tandem with the changing conditions, predicting changes as they occur. Saad’s team, along with their partners at the Institute of Science Tokyo, coined a term for such a thing: an internet of federated digital twins.

“What if everything is replicated? And they need to communicate, all these twins. How do you create an internet of digital twins?” asked Saad, who started to see this digital world as something potentially much more expansive.

He’d been thinking about the challenge of a mass-scale buildout of digital twins for existing wireless networks. But this flipped the question in his mind.

“Instead of saying, ‘We need digital twins for communication networks,’ now it’s, ‘We need communication networks to support digital twins of other things,’” he said.

To create this world effectively requires a unified platform that can seamlessly integrate digital twin systems across different domains. Increasingly large data volumes need scaling to match. Minimal latency of communication between the real and the digital world is crucial. And the system must be flexible, adaptive, and capable of responding to real-world challenges without any of the prior pieces falling by the wayside.

Another key component is identifying  what items actually need intelligence — people, other cars, dynamic objects — and what just needs to be replicated. By having an understanding of that, an object and its environment can communicate new information with one another as it develops.

“If you can predict something in the future and you know how the world will react with you, you’ll be able to know how to optimize your system now,” said Omar Hashash, a Ph.D. candidate in electrical and computer engineering working on the project.

Another key component to making digital twins work is edge computing, or localizing the digital environment in close proximity to the real one.  

“We can process data locally, reducing latency and ensuring rapid responses to changes in specific areas,” said Yu Tao, a member of the team from the Sakaguchi Lab at the Institute of Science Tokyo.

But for local, interconnected environments to exist all over will require a massive scaling. Platforms will need to be open to integrate across different systems. And all that will likely need the kind of computing power that AI provides.

“To reach a sustainable and scalable internet of federated digital twins solution that attracts industrial investment, our main focus is on building an open, standardized, interoperable platform that supports smooth integration across different industry systems,” said Tao.  

So, as Saad tells it, wireless will need AI to progress — but AI will also need wireless to do so in its own right. Developing in tandem, they have the opportunity to open up new possibilities for one another. 

With both Saad and Liu working under the same roof as Naren Ramakrishnan and the Sanghani Center for Artificial Intelligence and Data Analytics in Virginia Tech’s new Alexandria location, all the pieces are in play to drive a new generation of innovation.