Heterogeneous Statistical QoS Driven Collaborative Learning Based Energy Harvesting Over Full-Duplex Cognitive Radio Networks

Abstract

With the recent developments in energy harvesting (EH) technologies, mobile devices are able to support the wireless multimedia services by harvesting energy from various external sources. As one of the promising technologies to solve the energy scarcity problem, EH schemes have brought many new challenges due to the stochastic nature of the wireless channel and the harvested energy in supporting the statistical quality-of-service (QoS) provisionings. On the other hand, the full-duplex spectrum sensing (FD-SS) scheme has been designed to improve the spectrum efficiency while significantly enhancing the system performance over cognitive radio networks (CRNs). However, due to the unknown dynamics of the channel state information and the energy state information, it is challenging to design the efficient EH based power allocation policies for all the users with different operating modes while guaranteeing the statistical delay-bounded QoS constraints. To overcome the aforementioned problems, in this paper we develop the collaborative learning system model by choosing the optimal operation strategies and power allocation policies through learning from the EH process while satisfying the heterogeneous statistical delay-bounded QoS constraints over CRNs. In particular, we establish and analyze the FD-SS based EH system models over CRNs. Under the heterogeneous statistical delay-bounded QoS requirements, we formulate and solve the max-min fairness effective-capacity optimization problem for the battery-free EH based CRNs. Then, we apply the collaborative learning algorithm for deriving the optimal joint EH based mode selection and power allocation schemes. Also conducted is a set of simulations which evaluate the system performances and show that our proposed collaborative learning based EH scheme outperforms the other existing schemes under the heterogeneous statistical delay-bounded QoS constraints over CRNs.