Google LLC is getting closer to fulfilling its vision of enabling truly autonomous network operations with the launch of its latest artificial intelligence agents.

They’ll help telecommunications providers to build and maintain digital twins of their networks that can be used to predict their behavior under real-world conditions and test the impact of upgrades before they’re implemented. The updates, announced at MWC Barcelona, build on the release of Google’s Autonomous Network Operations framework last year, which serves as a blueprint for telcos to build self-healing, “zero-touch” networks powered by AI agents that can sense, reason and take autonomous actions without human oversight.

Google said its goal is to create networks that can identify, diagnose and fix problems on their own, without any human intervention, and its network digital twin is a key element of that vision. More than just a static map, it’s a dynamic, temporal graph that represents a network’s live physical and logical state, allowing telcos to understand its real-time performance and test how it acts under different conditions. Companies can use these digital twins to predict what problems a failure might cause, and also query their historical states to perform root cause analysis of earlier issues.

To accelerate adoption of its network digital twins, Google is releasing the source code for its telco data pipeline and data models on GitHub. Communications services providers will now be able to implement unified, industry-standard ontologies without having to perform manual schema mapping, making it much easier to build digital twins of their networks, Google said.

As for the new agents, they include a new data steward that’s tasked with automating data governance to ensure that digital network twins remain accurate and always reflect the current state of the real, physical network. In addition, it’s launching “autonomous network agents” designed to manage voice core networks that manage voice communication services, and operational support systems, which are used to orchestrate network operations.

Unlike traditional AI tools, these agents are designed to go beyond monitoring network conditions and take actions, such as independently rerouting traffic through the best available path in the event of an outage, or restoring call quality if it detects a drop in performance.

Google said a number of telco partners have embraced the chance to integrate more intelligence into their networks, including the likes of Deutsche Telekom AG and Vodafone Group Plc. In addition, it’s working with partners such as MasOrange S.L. and NetAI Inc. on a GraphML-based AI Operations project, and Nokia Corp. on a new network-as-code initiative that aims to make networks programmable via natural language prompts.

Image: SiliconANGLE/Meta AI