Learning to mine: integrating channel quality patterns for enhanced AI-assisted scheduling decisions in 6G multimedia networks

This paper addresses the growing complexity of 6G Radio Resource Management (RRM) due to the integration of
new multimedia services and sophisticated network functionalities. To handle complex RRM control issues, machine learning is recognized as a valuable tool for improving decision-making in scheduling. However, the effectiveness of intelligent scheduling methods can be limited without efficiently managing the large volumes of data produced by network measurements, such as Channel Quality Indicator (CQI) reports. To tackle this, we propose a low-complexity framework that compresses extensive CQI data into a compact, usable form for machine learning applications. The framework adopts a metaheuristic approach
with simulated annealing to perform clustering and identify
global CQI data centers. These centers are then used to train
a Radial Basis Function Network (RBFN), enabling efficient
classification of CQI data into distinct patterns. These patterns inform reinforcement learning-based scheduling decisions, which, in turn, achieve significant gains, up to 20% improvement, in fairness performance when scheduling multimedia consumers.