Abstract
In this paper, we investigate information and energy cooperation in cognitive non-regenerative two-way relay networks, where the multi-antenna secondary user (SU) transmitter harvests energy from primary users' (PUs) signals by the power splitting (PS) scheme to forward PUs' signals and transmit its own signals. Our objective is to design beamforming matrix and vector at the SU transmitter and PS factor to maximize achievable rate at SUs while maintaining achievable rate requirements at PUs subject to transmit power constraint at SU transmitter. We consider two scenarios that the SU transmitter knows perfect and imperfect channel state information of all links. For the former, we propose an optimal solution based on Charnes-Cooper transformation and 1-D search. We also propose a low-complexity suboptimal solution based on the zero-forcing (ZF) scheme and the algebraic norm-maximizing scheme. For the latter, channel uncertainties are modeled by a worst-case model. We propose to employ a ZF scheme and transform the problem into a semidefinite programming with the help of S-procedure and rank-one relaxation. The rank-one feasible solution is proposed to be recovered by the Gaussian randomization method. Simulation results demonstrate that our proposed designs outperform the beamforming scheme without energy cooperation.