6G-AINA

As 5G deployments get underway, the focus of wireless research is increasingly shifting towards the definition of 6G. In this process, we are witnessing a progressive introduction of sophisticated technologies and paradigm shifts such as Terahertz communications, ultra-massive and holographic radio, scalable cell-free MIMO networking, intelligent reflecting surfaces, interference management via non-orthogonal multiple access, massive random access, or highly-specialized slicing, network virtualization and disaggregation.

The development and deployment of these technologies entails complex management solutions to fulfill the stringent requirements (e.g., sub-ms latencies, Tb/s peak data rates, location accuracy at the cm level, etc.). Such complexity can only be addressed by introducing increasing levels of network automation to facilitate efficient resource exploitation.

In fulfilling such automation goals, the native integration of Artificial Intelligence (AI) and Machine Learning (ML) will play a pivotal role. Native integration of AI/ML refers to the inclusion of data-driven approaches from the onset, rather than as an add-on.

More specifically, the 6G-AINA project will focus on two main research areas from a physical and link layer perspective:

• AI/ML-driven air interface design and optimization
• the integration of localization and sensing capabilities into system definition

These correspond to two out of six major technology transformations identified in Nokia’s recent White Paper [1].

From a research perspective, the rising complexity of wireless systems also imposes severe constraints on deriving analytically tractable and computationally viable mathematical models (model deficit) that remain true to their actual physical behaviour. Besides, the optimization of the related objective functions is often not feasible, necessitating suboptimal solutions, often based on poor heuristics, leading to significant performance degradation (algorithm deficit). AI/ML happens to be particularly useful in such scenarios with a model deficit or algorithm deficit.

For these reasons, in 6G-AINA we advocate hybrid AI-driven/model-based approaches wherever this allows to leverage on years of engineering know-how developed to e.g., establish guarantees on QoS performance.

[1] H. Viswanathan and P. E. Mogensen, “Communications in the 6G era,” 2020, Nokia Bell-Labs White Paper.