Offline Power Allocation and Spectrum Sensing Strategy in Energy-Harvesting Cognitive Radio Networks

More Info
expand_more

Abstract

Energy-harvesting cognitive radio network has emerged as a solution to increase energy and spectrum efficiency. In this thesis, we propose short-term offline optimal power allocation algorithms for multi-user energy-harvesting cognitive radio networks considering interference between secondary users. Assuming finite and rechargeable batteries for secondary users and a time-slotted operation model, an off-line optimization problem is formulated so as to maximize the network throughput during finite time-period. To that aim, the design of a power allocation and the spectrum sensing strategy is required. Together with the inherent constraints imposed by the use of energy-harvesting devices, a collision constraint is also required to limit the probability of interference with the primary user and to guarantee the quality of service. Because of the intractability of the power allocation problem in the interference channel, we spilt the optimization task for two different size cognitive radio networks: a) 2-user network, and b) multi-user network (i.e. more than 2 users). The optimal algorithms are developed for a sharing single-frequency band, and a multi-band scenario for the two-user network. We derive the optimal solution following a two-step strategy in case of a 2-user energy-harvesting CR network. A suboptimal algorithm that entails reduced computational cost and performs very close to the optimal one is also proposed for the single-band sharing scenario. In case of multi-user energy-harvesting cognitive radio networks, a SQP-based sub-optimal algorithm is derived for single-band sharing scenario. Besides, a optimal solution is proposed for a CR network applying interference cancelation techniques sharing the single band. At last, we derive the optimal power allocation strategies for the multi-user multi-band sharing scenario. Simulation results of the optimal (and suboptimal) solutions outperform those achieved by a random or priority best-user power allocation algorithms for the AND and OR fusion rules.