Power-efficient frequency allocation algorithms for self-organized networks

In this paper we compare three novel schemes for dynamic frequency allocation and transmission power assignment for cognitive radio networks. These spectrum allocation schemes are based on reinforcement learning and power optimization procedures. We propose the so-called RL+Power and RL+Power Coupled schemes and compare them with an existing constant power scheme. Simulation results show that the RL+Power Coupled scheme improves the QoS of the system by up to 5 dB in terms of SINR and by up to 10% in terms of user satisfaction when compared against the two other schemes in a high cell load environment.