Towards Cognitive Service Delivery on B5G through AIaaS Architecture
Larissa F. Rodrigues Moreira, Rodrigo Moreira, Flávio de Oliveira Silva, André R. Backes
TL;DR
This paper addresses the limitations of the existing NWDAF in enabling widespread cognitive service delivery within B5G/6G networks. It proposes an evolved NWDAF (e-NWDAF) architecture that incorporates AI4Net and Net4AI concepts, enabling AIaaS and third-party cognitive services through OSS/BSS interfaces and AI-Agents microservices, with TrainingSlice and InferenceSlice for flexible AI deployment. The authors provide a architectural framework and a set of research directions to realize distributed NWDAF capabilities, emphasizing practical pathways toward AI-driven core-network services. The work lays groundwork for integrating cognitive services at the network core, potentially enabling new NetApps and AIaaS models in future mobile networks, while acknowledging the need for empirical validation and benchmarking.
Abstract
Artificial Intelligence (AI) is pivotal in advancing mobile network systems by facilitating smart capabilities and automation. The transition from 4G to 5G has substantial implications for AI in consolidating a network predominantly geared towards business verticals. In this context, 3GPP has specified and introduced the Network Data Analytics Function (NWDAF) entity at the network's core to provide insights based on AI algorithms to benefit network orchestration. This paper proposes a framework for evolving NWDAF that presents the interfaces necessary to further empower the core network with AI capabilities B5G and 6G. In addition, we identify a set of research directions for realizing a distributed e-NWDAF.