Читать книгу Shaping Future 6G Networks - Группа авторов - Страница 27

As discussed previously, the disaggregation principle with loosely coupled network functions (including in the RAN) will be at the core of the forthcoming 6G networks. This allows the dynamic orchestration of network functions and thus adaptation of the network to specific service requirements and network conditions in an end‐to‐end manner. In this context, monitoring and management of data will become the new fuel in networking. Chapter 13 addresses major development trends toward the definition of a data‐layer‐oriented network. It presumes the extraction of a large amount of data, its exchange across different elements, the generation of insight, and its immediate application as customized configurations across the system. A new type of network optimization is obtained based on user behavior instead of the one of the systems complimented by a large level of native automation.

Chapter 14 discusses opportunities given by the adoption of ML at the edge of wireless networks. The authors describe the latest in ML (e.g. Kernel‐based learning, deep learning, and reinforcement learning) and suggest application domains within the radio network exploiting the available domain knowledge, such as robust traffic prediction for energy optimization or optimized localization of end systems and beamforming optimization. They proclaim that ML will be the heart of future 6G architectures. Chapter 15 expands on the previous chapter by looking specifically at the adoption and standardization of AI/ML for secure, automated end‐to‐end slice orchestration and management, both at the edge and at the core of 6G mobile networks. The authors describe the rising use of AI/ML across the control protocol stack, investigating in particular the use of federated learning for optimal communication. They also introduce the challenges for global management of 6G systems that should perform smart resource management, automatic network adjustment, provisioning, and orchestration and rely on real‐time data insights to optimize network performance. Standardization should play a key role to achieve this goal, as already started with OpenRAN.