RL-Based Mobility Load Balancing for Energy Efficiency in Multi-Band Cellular Systems

Abstract

This paper presents a mobility-aware framework for energy-efficient load balancing in homogeneous multi-band cellular networks. The proposed approach combines a newly proposed user-association algorithm, termed SIRRA-HO (SINR-Ratio-Based Resource Allocation under Handover Constraints), with a Proximal Policy Optimization (PPO)-based power-control scheme for high-frequency bands. By prioritizing users with larger discrepancies between their best and second-best feasible serving bands, SIRRA-HO mitigates the loss caused by suboptimal association and improves mobility stability. The framework explicitly captures time-varying user locations, realistic handover feasibility, and dynamic propagation conditions. Simulation results show that the joint scheme achieves more than 24% energy-efficiency gain relative to the baseline, while maintaining low dropped-user levels. They further show that RL-based power adaptation provides the dominant energy-efficiency improvement, whereas the proposed SIRRA-HO algorithm contributes mainly through reduced handover activity and improved mobility robustness. Overall, the results highlight the effectiveness of combining the proposed handover-aware user association with RL-based power control for next-generation cellular networks.