Distributed AI-native Architecture for 6G Networks

Abstract

How to achieve native intelligence in 6G networks has attracted widespread attention and discussion in the communication field. In the core network, NWDAF (Network Data Analytics Function) is introduced as an AI (Artificial Intelligence) service NF (Network Function) to provide intelligence analysis, such as supporting network load and congestion prediction, UE behavior, and service experience statistics. However, there are still problems such as the lack of AI capability of NFs (Network Functions), weak cross-domain collaboration, and little support for distributed learning methods, etc. This paper presents our solutions on native intelligence in 6G networks, including the overall distributed network architecture of native intelligence, emphasizing the capabilities of each intelligent node and the collaboration among nodes, and further proposes some thoughts and potential solutions from the perspective of cross-domain coordination, and the combination of distributed learning and the network.