AI-Based Digital Twins for Multi-intent 6G Networks – an Architectural Perspective

RajaGopalan Rajappa, CTO of Tata Elxsi, in his keynote at 5G O-RAN alliance next generation Research Group (ngRG) workshop on Feb 22, 2024, held at Athens in Greece emphasized the significant role of AI Based Digital Twin for Evolutionary architectural framework of 6G Architecture.

Participants were introduced to Tata Elxsi, its vertical industry expertise and its long wireless experience of over 20 years as a journey through its product development of CDMA, 1xEVDO base station, 3G NB stack, WiMAX, LTE IPS, its transition into software driven SDN, NFV, network orchestration, automation and industry specific use cases such as Connected Vehicle, Connected Office, Homes and Connected Factories in a brief overview.

The talk focussed on how AI and Digital Twin based 6G Architecture relies heavily on data and data driven approach that imposes fundamental change in how the 6G technology evolves from its previous generation of wireless technologies. The cloud based 5G SA architecture enabled the Continuous integration and Continuous deployment whereas AI based 6G pushed the frontiers of software centricity further to microservices/container-based software model into RAN where ML based Channel coding, Channel estimation, Interference Mitigation, Cross layer scheduler optimization can run as containers and microservices with CI/CD linked to them. 6G AI native technology evolution is shown with 4G and 5G in Figure 1 below.

Figure 1: (r)Evolution - Al Based Digital Twin 6G

It is emphasized that data, data sets, and KPIs to be standardized across the RAN, Core, Transport and Service domains, along with data sharing and federation among the ecosystem to speed up the learning cycles of AI/ML based models across the AI native 6G technology stack for the faster maturity and adoption.

The talk provided an overview of how data-driven AI/ML algorithms on Digital Twin can help verify the complex ML model algorithms for various aspects of 6G network, such as O-RAN physical layer, Transport routing, network slicing on demand, end to end network capacity planning and estimation, and HetNet dense deployments in 6G Terahertz frequencies, by using What-If experiments on Digital Twin as a test platform. It also presented the advantages of Digital Twin as a powerful hybrid Phy-Gital (Physical+ Digital) test environment for complex deployments that involve Reconfigurable Intelligent Surface, network operations, and remediation procedures. It is illustrated below in Figure 2.

Figure 2: Digital Twin - 6G Network Deploy/Innovate/Operate accelerator (many archetypes)

The speaker shared his opinion that the innovation cycle will be improved by making the data open and standardized, and allowing everyone to access and use it. This will enable academia, startups, and new players to participate and accelerate the 6G Flywheel cycle that relies on Open sources, Open systems, Open Interfaces and Open Democratized data. The cycle involves constant learning, innovation, integration, and deployment.

The Digital Twin, which relies heavily on data and artificial intelligence, presents significant challenges. They are listed along with possible solutions in Figure 3.

Figure 3: Digital Twin – Challenges & Solution

Briefly, Tata Elxsi’s approach for 6G Autonomous network technology stack and its components, which are based on open systems, were presented. These include Tata Elxsi’s NEURON, TEDAx, LEXi (Digital Twin) & TETHER (Intelligent Telemetry & Data collection). It is illustrated below in Figure 4.

Figure 4: Tata Elxsi’s Autonomous Network Stack Architecture for 6G

The speaker concluded by discussing about various business models for Telco ecosystem towards 6G network such as RAN as Service, Data as a Service, ML Model as a Service, AI Infrastructure as a service in addition to Network as a Service. Further to that, an earnest appeal was made to SDO’s towards lobbying the regulatory bodies to offer the spectrum as a service and to recommend for the differential commercial models for the spectrum when used for education, rural health care, rural welfare & upskilling in order to create the larger societal impact that 6G promises in its vision. It is summarized in Figure 5.

Figure 5: AI Based Digital Twin for 6G Network – New Business and Operating Model