This Bottom Mount paper presents an algorithm that supports the dynamic spectrum access process in cognitive radio networks by generating a sorted list of best radio channels or by identifying those frequency ranges that are not in use temporarily.The concept is based on the reinforcement learning technique named Q-learning.To evaluate the utility of individual radio channels, spectrum monitoring is performed.In the presented solution, the epsilon-greedy action selection method is used to indicate which channel should be monitored next.The article includes a description of the proposed algorithm, scenarios, metrics, and simulation results showing the correct operation of the approach relied upon to evaluate the utility of radio channels and the epsilon-greedy action selection Tongue Depressors method.
Based on the performed tests, it is possible to determine algorithm parameters that should be used in this proposed deployment.The paper also presents a comparison of the results with two other action selection methods.