Scalable optimization of small cell networks using reinforcement learning

In this research work, the authors propose a novel approach through a reinforcement learning-based algorithm which aims to optimize the performance of ultra-dense small cell networks in 5G and beyond systems. The RL agent that has been proposed is responsible for controlling the transmission power, spectrum overlay, and load balancing in a quite selective manner with the purpose of cutting down the frequency of delays and interferences while at the same time improving the quality of the signal and the energy efficiency of the system. Furthermore, a distributed architecture is implemented, which mitigates communication burden and increases scalability. The simulation results demonstrate that the proposed RL-based scheme reduces average latency by approximately 8–12%, improves user throughput by 10–18%, decreases power consumption by 38%, and enhances SINR by 5–10% compared to baseline methods, thereby outperforming Fixed Power Allocation, Greedy Allocation, and Proportional Fair scheduling schemes. Furthermore, the system’s performance stays stable even when the number of user devices goes up from 50 to 300, thereby demonstrating stable performance under the evaluated network scenarios. Therefore, these results indicate that RL-based optimization is a promising approach for designing intelligent, adaptive, and energy-efficient next-generation small-cell networks within the evaluated simulation scenarios.