Distributed sensing and access in cognitive radio networks

We consider an ad hoc network of secondary users searching for idle frequency bands in a spectrum consisting of multiple channels. In each slot, a secondary user chooses one channel to sense and decide whether to access based on the sensing outcome. A sensing strategy for intelligent channel selection is crucial to track the rapidly varying spectrum opportunities. In an ad hoc network without a central controller or common control channels, a secondary user can only resort to its local observations in the decision making. The tradeoff is between choosing the channel most likely to be idle and avoiding other competing secondary users. We show that the problem can be formulated as a decentralized Partially Observable Markov Decision Process (POMDP). A suboptimal randomized sensing policy is then proposed. This policy effectively addresses this design tradeoff and offers significant improvement in network throughput over the optimal single-user design.