Relay selection Algorithm for wireless cooperative networks: a learning-based approach

Relay selection in cooperative communication is a crucial task for achieving the spatial diversity since the improper relay selection can decrease the overall capacity of the network. In this study, the authors use a reinforcement learning technique, called as Q-learning (QL), to solve the relay selection problem. They propose a 'QL-based relay selection algorithm' (QL-RSA) for wireless cooperative networks that maximises the total capacity of the network. QL-RSA receives the reward (feedback) in terms of the capacity by learning a multi-node amplify-and-forward cooperative environment with time-varying Rayleigh fading channels. The advantages of QL-RSA are that it is less complex, requires less channel feedback information and it is distributed in a multiple-sources environment as it provides each source a self-learning capability to find the optimal relay without exchanging information with other source nodes.