Radio Access Scheduling using CMA-ES for Optimized QoS in Wireless Networks

In this paper, we explore a radio access scheduling problem, TimeFreqResourceAllocation-v0, proposed by Nokia Bell Labs in its open-source framework Wireless Suite. We elaborate on a learning-based black-box methodology using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) as a candidate solution for QoS-aware scheduling. The proposed approach achieves the best score performance compared to either baseline methodologies (e.g., Round Robin, Proportional Fair, etc.) or common Deep Reinforcement Learning approaches.