The use of machine learning and artificial intelligence in test tools isn’t new. But it is seeing new innovation and strategies, in order to meet operators’ increasingly complex testing needs as they seek to bring more AI features into their networks.
“AI is great, but how do you validate that it’s going to be integrated successfully?” asks Sameh Yamany, CTO at VIAVI Solutions. Assessing AI capabilities in the context of a telecom network—whether it is working, whether it is accomplishing its desired task as expected, whether it is more effective than other tools—are not exactly simple, pass/fail scenarios.
Enter the digital twin, which builds upon the long-standing strategy of emulation: Replicating real-world conditions and impairments such that you have essentially created an accurate digital replica of an element of a telecom network, a discrete portion of a network or even (maybe someday) an entire network, synchronized in real-time with data feeds from the actual network itself. The VIAVI RAN Digital Twin creates a virtual version of the operator’s RAN. It combines synthetic traffic with real data from probes, sensors, and third-party sources like traffic and weather reports. TeraVM AI RSG is the engine behind this solution, simulating RAN behavior with high accuracy, modeling different network setups, configurations, mobility patterns, and traffic flows.
While there isn’t a common definition of what makes a system a digital twin, or the point at which advanced emulation crosses into digital twin territory, the building blocks of that level of digital re-creation are already here, as a result of the convergence of powerful computing and data processing capabilities; large amounts of digital data coming from software-centric networks and the visibility solutions which observe them; and the telecom industry’s push toward Open RAN and applying artificial intelligence through the AI-assisted RAN Intelligent Controller.
VIAVI, with its history of sophisticated emulation capabilities, leaned into that convergence as a jumping-off point for developing digital replicas of the building blocks of physical telecom systems: User equipment; the core network; the RAN; Non-Terrestrial Network infrastructure. Each portion was flexible enough to be used for validation, for building out a particular use case across the network, or for testing a specific part of the network such as Open RAN implementations and RIC effectiveness.
It turns out that once you can generate incredibly detailed, realistic RAN data that can both train x/rApp algorithms and later, be used to put them through their paces, then combine that with the ability to validate how well those x/rApps handle things like traffic steering, balancing energy efficiency vs. user experience, and mitigating conflict between apps with dueling purposes—you are awfully close to a full-fledged RAN digital twin.
VIAVI frames the AI RSG solution as an AI enabler within the network, focused on data generation. It can simulate thousands of UEs and cells per reference server. Users can import real-world maps with streets, buildings, and network configurations to create UE profiles that simulate movements, handovers, and resource requests. Traffic demand profiles can also be generated, as can various radio conditions through propagation models, channel emulation and mobility modeling. Which sounds a lot like what one would expect of a digital twin, in that you can turn algorithms loose and see what happens, with high confidence that the outcomes will be realistic.
“RAN digital twins are shaping what the overall network digital twin will look like,” Yamany said. “We’re making sure that any AI algorithms that are put into that network are actually making the right decisions. You need something to start off with, to see how those decisions will be made and what impact that’s going to have on the network, before you’re going to let that live on the network.” The twin is not just allowing you to test the expected e.g. software updates, security patches etc., but enabling you to test the unexpected, the “what if” scenarios. A digital twin allows safe experimentation to cover those events that service providers are concerned about in the safety of a lab.
The industry is still a ways off from holistic, live network digital twins that are fully synchronized and real-time representations of all aspects of the network, Sameh Yamany notes; those are likely to come in stages.
But one stage is arguably already here, and customers are already using it. Combined with VIAVI’s TM500 and the company’s other emulation options and NITRO Location Intelligence, the AI RSG is already enabling VIAVI Solutions to not only provide crucial pre-deployment insights for customers preparing for 5G-Advanced, but also support 6G development through replicating experimental radios and playing out the possible behaviors of technologies such as joint communications and sensing (JCAS).
To learn more about VIAVI Network Digital Twin solutions, check out: https://www.viavisolutions.com/network-digital-twin
