Distributed selection of sensing nodes in cognitive radio networks

The introduction of cognitive radio networks is dependent on the reliable sensing of higher priority emitters in their geographic areas. A large number of sensors requires the exchange of a substantial amount of information; therefore a subset of secondary nodes must be selected that provides sufficient geographic coverage. In this paper, a new approach to select the reporting nodes is introduced based on the affinity propagation (AP) clustering algorithm. The AP algorithm operates in a distributed fashion and requires only link quality information between neighbouring nodes. Simulations incorporating pathloss and correlated shadow fading models show that the sensing reliability is significantly better than random node selection, and is comparable to a commonly-used centralized algorithm that requires knowledge of the positions of all nodes.