Active queue management in 5G and beyond cellular networks using Machine Learning

This paper proposes a state-of-the-art framework for adapting Active Queue Management (AQM) in 5G and beyond cellular networks with disaggregated Radio Access Network (RAN) deployments. While existing AQM algorithms effectively mitigate bufferbloat in monolithic RAN deployments, their potential in disaggregated ones remains largely unexplored. This gap particularly relates to AQM algorithms relying on communication between layers distributed across distinct network entities to operate. Our research explores the current literature on AQM, identifies the gaps regarding disaggregated deployments, and introduces a comprehensive framework that employs Artificial Intelligence (AI) and Machine Learning (ML) within the RAN Intelligent Controller (RIC) for adapting AQM in such deployments. We evaluate our novel solution on a previously proposed AQM algorithm which requires cross-layer communication, using OpenAirInterface5G (OAI5G) to deploy a disaggregated RAN and a connected User Equipment (UE) that experiences realistic network conditions, including noise and mobility. Finally, we assess its accuracy through the Quality of Service (QoS) achieved for our disaggregated deployment on the NITOS testbed.