This paper focuses on the use of Opportunistic Networks as a candidate solution for extending the coverage of cellular networks when providing high bit rate data services. It is based on establishing a device-to-device (D2D) radio link with another mobile that can provide the connectivity to the infrastructure. Specifically, the paper proposes a novel cognitive solution for the joint selection of node and spectrum to be used in the D2D radio link. It makes use of learning-based mechanisms for supporting the decision-making process taking into account the context information and the application requirements. Simulation results reveal that the performance of the proposed approach is very close to the optimum one. At the same time it is capable of adapting to changes in the scenario, such as the appearance/disappearance of candidate nodes as well as the variability in the interference conditions in the different bands.

Peer Reviewed