Power allocation in cognitive radio networks using distributed machine learning

In cognitive radios systems, the sparse assigned frequency bands are opened to secondary users, provided that the aggregated interferences induced by the secondary transmitters on the primary receivers are negligible. In this paper, a decentralized Q-learning power allocation algorithm is proposed to maximize the Signal on Interference-plus-Noise Ratio (SINR) at the secondary receivers while meeting the primary protection constraint. Numerical simulations allow to compare the performance of the proposed algorithm with the performance of the optimal centralized power allocation scheme in the scenario of multiple independent secondary cells located on the border of a primary network protection contour. The impact of the definition of the learning algorithm cost function on the secondary user fairness and the system performance is discussed. The optimal execution frequency of the algorithm is obtained and different strategies for the exploration of new solutions are compared.