Multi-band RF Energy and Spectrum Harvesting in Cognitive Radio Networks

Energy harvesting (EH) from Radio Frequency (RF) sources is considered a promising solution for future wireless technologies in terms of efficiency and battery lifetime, specially for applications that do not consume a lot of energy such as Internet of Things (IoT) devices. This paper investigates a multiband EH schemes under cognitive radio interweave framework. All secondary users are considered as EH nodes that are allowed to harvest energy from multiple bands of RF sources. A time switching protocol, by which the SUs switch over time between sensing and energy harvesting is used by the SUs. A win-win framework is proposed, where SUs can sense the spectrum to determine whether the spectrum is busy, and hence they may harvest from RF energy, or if it is idle, and hence they can use it for transmission. Only a subset of the SUs can sense in order to reduce sensing energy, and then machine learning is used to characterize areas of harvesting and spectrum usage. More specifically, the SUs in the former region can harvest from RF energy, while the SUs in the latter region can access the spectrum. We formulate an optimization problem that jointly optimize number of sensing samples and sensing threshold in order to minimize the sensing time and hence maximize the amount of energy harvested. A near optimal solution is proposed using Geometric Programming (GP) to optimally solve the problem in a time-slotted period. Finally, an energy efficient approach based on multi-class Support Vector Machine (SVM) is proposed by involving only training SUs instead of all SUs.