Application of multi-armed bandit algorithms for channel sensing in cognitive radio

We apply a multi-armed bandit algorithm for channel sensing in a cognitive radio system. In order to find a channel whose utilization rate is low, we formulate several problems in cognitive wireless systems as the multi-armed bandit problems, which is the problem to select the best arm providing rewards with the highest rate within smaller sensing time. There are several effective algorithms to solve the multi-armed bandit problems. In this paper, we introduce the e-greedy algorithm and apply it to selection of a channel with lowest utilization ratio. We also extend the proposed scheme for a frequency hopping system, since the frequent use of the vacant channel can be corresponded to the increase of rewards by multi-armed bandit algorithms. In order to evaluate the performance of the proposed scheme, we use real measured data in 2.4GHz band. The results show that the proposed scheme improves the performance of the communication systems, by efficient sensing of the channels with lower utilization rate.