Power Control Based on Deep Q Learning in Heterogeneous Networks
Yingqi Zhao, Jin Jin, Tingting Yang · 2024
Heterogeneous network is a promising solution to meet higher network capacity requirements and faster transmission rates. Under the spectrum-sharing strategy, the interference problem among user devices is a major factor affecting the network’s performance. Power control methods employing deep reinforcement learning are extensively utilized for interference management purposes. A previous study formulated a multi-agent power control algorithm utilizing deep reinforcement learning to tackle the sum-rate optimization problem within a fixed base station deployment scenario in heterogeneous networks. Specifically, the aforementioned research exclusively examined a specific scenario wherein the femto base stations are situated within the central coverage area of the macro base station. Inspired by the research, this paper considers the variation of femto base stations’ location and explores the impact of base station distribution on the sum-rate using deep Q learning algorithm. Simulation results show that the average sum-rate enhances with the increase of the sum distance between the macro base station and the femto base stations.