| NOTAS: | "This document provides a broad view of the functional aspects that needs to be considered for the incorporation of native and cross domain AI into next generation networks. It begins with a brief overview of the current status of AI in global standards organizations, including 3GPP, O-RAN and ETSI-ZSM. The report provides concise definitions of the terms native and cross domain AI in the context of wireless networks and then goes on to discuss the impact of AI on the architecture. The challenges of ingesting large amounts of disparate data across multiple layers, and the corresponding requirements on data modeling, formatting and representation are discussed. A unified data ingestion model is emerging as a key requirement. The importance of distributed and edge intelligence to solve the complex multi-layered issues in wireless networks is emphasized, along with the notion of trustworthiness in such a distributed architecture. Enablers for large scale distributed intelligence, including HPC platforms and accompanying software platforms including open-source, are discussed. The emerging paradigm of intent-driven management, and its interplay with AI/ML are described. The necessity to have collaborative AI across disaggregated RAN and between RAN and CN are discussed.This research report is the first attempt in O-RAN next Generation Research Group (nGRG) to survey the landscape with respect to AI/ML as it applies to next generation networks, and provides a foundation based on which several further explorations into each of the highlighted areas can be initiated." |
