SDAC: An Architectural Enhancement to Enable Artificial Intelligence in 5G Systems

Abstract

This paper explores two new ideas to enable Artificial Intelligence (AI) and Machine Learning (ML) in 5G Systems, which is an essential need for modern networks, allowing users to utilize multiple access technologies (cellular and Wi-Fi) simultaneously. The first idea proposes to connect a meddler component so-called “Stack Data Analytics Coordinator” (SDAC), with each radio protocol stack (e.g., 5G-NR or Wi-Fi) at both user equipment (UE) and access network nodes (e.g., gNB and Wi-Fi AP/Controller). SDAC acts as a coordinator between data providers (which could be in a protocol stack) and analytics providers (which could be anywhere in the 5GS and UE). However, if an analytics consumer is within the protocol stack (e.g., at the MAC layer), then SDAC allows an analytics provider to be operating close to the protocol stack (e.g., at the same box), minimizing end-to-end communication latency between these components. Furthermore, SDACs allow UE, WLAN (Wireless LAN), RAN (Radio Access Network), and Core Network (CN) to directly interact with each other and exchange statistics, measurements, and analytics in a flexible manner (i.e., fast and with low overhead). Hence, they facilitate the AI/ML deployments within UE, RAN, WLAN, and CN. To realize SDAC, several new interfaces are defined, including Napp, Nsdac, Nwifi, and N5g. The second idea extends the network data analytics services concept, currently standardized in the 5G Core, into UE, RAN, and WLAN environments. This service expansion unifies the deployment of AI/ML techniques and also the way in which data and analytics should be stored and retrieved within UE, RAN, WLAN and CN. This way, e.g., an SDAC residing at RAN, close to a gNB, can interact with Network Data Analytics Functions (NWDAFs) operating in RAN, UE, and other locations.