Treffer: Machine Learning-Driven Active Queue Management for Latency Optimization in 5G and Beyond Networks.

This research introduces a cutting-edge Active Queue Management (AQM) architecture for 5G and future cellular networks with decentralised RANs. Traditional AQM algorithms function well in monolithic RANs but badly in partitioned protocol stacks. The disaggregation delay complicates cross-layer coordination. By adding ML and AI modules to the RAN Intelligent Controller (RIC), the proposed architecture allows latency-aware AQM operations. This technique is tested in realistic situations including user mobility and channel noise using the Dynamic RLC Queue Limit (DRQL) algorithm. The NITOS testbed was installed and tested on OpenAirInterface 5G (OAI5G) to ensure experimental fidelity under changing network conditions. The AI-assisted architecture reduces bufferbloat in complicated disaggregated deployments by maintaining high QoS and lowering latency. This work is a key step towards AI-powered queue management in next-gen mobile platforms. [ABSTRACT FROM AUTHOR]