Dynamic spectrum access with learning for cognitive radio

We study the problem of cooperative dynamic spectrum sensing and access in cognitive radio systems as a partially observed Markov decision process (POMDP). Assuming Markovian state-evolutions for the primary channels, we propose a greedy channel selection and access policy that satisfies an interference constraint and also outperforms some existing schemes in average throughput. When the distribution of the signal from the primary is unknown and belongs to a parameterized family, we develop an algorithm that can learn the parameter of the distribution still guaranteeing the interference constraint. This algorithm also outperforms the popular approach that assumes a worst-case value for the parameter thus illustrating the suboptimality of the popular worst-case approach.