A Distributed Online Learning Method for Opportunistic Channel Access with Multiple Secondary Users

In this paper, we investigate the problem of distributed dynamic spectrum access with multiple secondary users (SUs), where the availabilities of licensed channels are unknown to SUs. If multiple SUs access the same channel simultaneously, none of them would transmit successfully because of collision. As a result, each SU has to learn unknown channel statistics and coordinate with others based on its local observations. We develop an online learning based channel access method, which identifies the best channel quickly by using Thompson sampling and orthogonalizes SUs on different channels efficiently without prior information. Numerical results show that the proposed method achieves the highest probability of identifying the best channel compared to the existing ones.