ARTIFICIAL INTELLIGENCE ARCHITECTURE RESEARCH FOR 6G COMMUNICATION INFORMATION NETWORKS

Abstract

Technologies for building an artificial intelligence network architecture for telecommunication access networks, taking into account a large amount of data related to the operation and management of the network, user activity, the process of scanning the environment and the operation of end devices, have been studied. Features of the design of the new 6G system are considered, in particular, the effective organization of data coming from completely different areas and their management, taking into account privacy protection.

It was determined that the fundamental architectural difference between 5G networks and 6G networks is the built-in support for artificial intelligence in the 6G network. The architecture of the segment of the information communication network with built-in AI was studied, which is embodied in three business models: infrastructure as a service; platform as a service; artificial intelligence as a service. AI services running on this innovative infrastructure will bring many benefits, namely: a shift from global AI to local AI, as from a nationwide network perspective, centralized training is characterized by high cost, due to the fact that it involves collecting and sending data across the network to the central object.

It is planned to develop a task-oriented communication solution that covers four main aspects: task management, resource management/scheduling during work, data management and connection management. At the same time, from an architectural point of view, it may make sense to introduce new network services and APIs for task management that implement the definition, execution, and management of tasks throughout their lifecycle. A real-time scanning model with high-precision localization and tracking of user movements by network services is proposed.

A study of deep edge computing using the capability of artificial intelligence at the RAN level was performed. Opportunities to optimize resource planning and reduce interference while supporting AI based on 6G mobile networks are considered. Researched models to support the structure of AI and the resulting potential requirements for the mobile communication system are a key asset of the artificial intelligence industry. As the first wave of AI services are more focused on business-to-consumer (B2C) applications, end users are the direct data sources. The dependence of the implementation of deep learning (such as federated learning) on the main functional parameters of the communication system, i.e. throughput and delay, is determined. At the same time, the network system architecture can influence AI training and its logical results.