Applicability Domains Of Machine Learning In Next Generation Radio Access Networks

The Radio Access Network (RAN) is envisaged to undergo a significant transformation in the context of 5G and beyond mobile communications systems. One of the driving forces behind this transformation is the applicability of Machine Learning (ML) techniques. Taking as a reference the high-level architecture for a next generation RAN proposed by the Open RAN Alliance, this paper identifies the applicability domains where ML techniques can play a relevant role. For each domain, namely radio physical layer processing, Medium Access Control (MAC) scheduling, near-real time Radio Resource Management (RRM), RAN data analytics and RAN operational automation, the paper discusses the specific functionalities that can benefit from the application of ML and analyses the key considerations and challenges that need to be addressed when developing ML-based solutions, given the particular characteristics of the mobile radio environment.