Across Europe and internationally, research organizations are exploring new ways to accelerate AI innovation in optical transport networks while maintaining strict control over sensitive network data. As part of this effort, Adtran is collaborating with Fraunhofer HHI, Japan’s NICT and Trinity College Dublin to support a federated testbed approach – the Optical Testbed Data Space (OTDS) – that enables partners to develop and validate machine learning models safely, efficiently and across multiple domains.
A core focus of the work is data sovereignty, ensuring that operators, vendors and research partners retain full control over how optical network data is shared and used. Rather than pooling raw data, the approach relies on governed, policy-controlled access that respects ownership and confidentiality.
This work is being demonstrated at leading research events, including this month at the OFC Demo Zone, where the partners are showcasing how different optical testbeds can share selected telemetry using well‑defined policies and modern dataspace technologies.
A shared approach to data‑sovereign AI development
Optical networks generate valuable telemetry that can help improve failure prediction, performance optimization and overall service resilience. But much of this data cannot be shared freely. Operators have understandable concerns about data ownership, network visibility and confidentiality, while vendors face similar questions about exposing equipment performance.
To support this, the partners use a dataspace approach that allows each organization to control how its data is accessed. This is built on open‑source components from the Eclipse Foundation and follows the internationally recognized International Data Spaces model for governed data sharing.
This framework allows each participant to maintain full control over their data while enabling tightly governed access for research. Partners can define which data is shared, at what level of granularity and under what conditions.
This capability is particularly important in Europe, where digital autonomy and data sovereignty are recognized priorities for future network innovation. At the same time, the approach remains open and international, supporting collaboration with European partners alongside colleagues in Japan and beyond.
How Adtran supports the federated testbed
We’re contributing both live optical infrastructure and digital‑twin models to the federated testbed. Our FSP 3000 optical technology, deployed within the Fraunhofer HHI testbed, provides real optical signals and network behavior for cross‑domain experimentation, helping ensure the federated setup reflects practical, multi‑vendor environments.
Digital twin models and simulated networks
In addition to physical equipment, Adtran is contributing digital‑twin models of key components, particularly optical amplifiers, created using internal simulation environments. These models allow realistic behavior to be shared without exposing sensitive performance data. As federated data sharing matures, these models will help partners evaluate new AI techniques and test algorithms under controlled conditions.
Linking independent testbeds through governed data exchange
The collaboration uses a multi‑domain reference architecture that links the testbeds at HHI, NICT, Trinity College Dublin and Adtran. Each domain operates independently, with dataspace connectors managing governed access to selected telemetry and models across the testbeds.
