DBLA: Distributed block learning algorithm for channel selection in Cognitive Radio Networks

In this paper, we study the distributed channel selection problem in a cognitive network. We consider a time varying channel environment where secondary users do not have any prior knowledge of primary transmission and independently learn about existing channels without any explicit communication with other users. To solve this problem, we address the impact of switching between channels during the learning process that is mostly ignored in literature. We propose a scalable distributed block learning algorithm that minimizes the switching cost, adapts to time varying channel conditions, and achieves logarithmic regret. Simulation results show that our algorithm performs significantly better than the existing ones.