Channel Exploration and Exploitation with Imperfect Spectrum Sensing for Multiple Users

In this paper, the fundamental problem of multiple secondary users (SUs) contending for opportunistic spectrum sensing and access over multiple channels in cognitive radio networks is investigated, when sensing is imperfect and each SU can access up to a limited number of channels at a time. For each channel, the busy/idle state is independent from one slot to another. The availability information of channels is unknown and has to be estimated by SUs during channel sensing and access process. Learning loss, also referred as regret, is thus inevitable. To minimize the loss, we model the channel sensing and access process as a multi-armed bandit problem, and contribute to proposing policies for spectrum sensing and access among multiple SUs under both centralized and distributed framework. Through theoretical analysis, our proposed policies are proved with logarithmic regret asymptotically and in finite time, and their effectiveness is verified by simulations.