Randomized Spectrum Access in Cognitive Radio Networks with a Large Number of Cognitive Users

In this paper, we study the spectrum access problem in cognitive radio networks with the objective to maximize the expected throughput of cognitive users. We show that the performance of a uniform access policy that uniformly selects one channel to sense, independent of the primary users' behaviors on the channels, is close to that of the optimal randomized access policy when there are a large number of contending cognitive users. We also characterize the throughput taking the energy constraint of cognitive users into account, and present a unified PHY/MAC design to improve the throughput.