Practical Design for Multiple-Antenna Cognitive Radio Networks with Coexistence Constraint

Pin-Hsun Lin, Gabriel P. Villardi, Zhou Lan, Hiroshi Harada

In this paper we investigate the practical design for the multiple-antenna cognitive radio (CR) networks sharing the geographically used or unused spectrum. We consider a single cell network formed by the primary users (PU), which are half-duplex two-hop relay channels and the secondary users (SU) are single user additive white Gaussian noise channels. In addition, the coexistence constraint which requires PUs' coding schemes and rates unchanged with the emergence of SU, should be satisfied. The contribution of this paper are twofold. First, we explicitly design the scheme to pair the SUs to the existing PUs in a single cell network. Second, we jointly design the nonlinear precoder, relay beamformer, and the transmitter and receiver beamformers to minimize the sum mean square error of the SU system. In the first part, we derive an approximate relation between the relay ratio, chordal distance and strengths of the vector channels, and the transmit powers. Based on this relation, we are able to solve the optimal pairing between SUs and PUs efficiently. In the second part, considering the feasibility of implementation, we exploit the Tomlinson-Harashima precoding instead of the dirty paper coding to mitigate the interference at the SU receiver, which is known side information at the SU transmitter. To complete the design, we first approximate the optimization problem as a convex one. Then we propose an iterative algorithm to solve it with CVX. This joint design exploits all the degrees of design. To the best of our knowledge, both the two parts have never been considered in the literature. Numerical results show that the proposed pairing scheme outperforms the greedy and random pairing with low complexity. Numerical results also show that even if all the channel matrices are full rank, under which the simple zero forcing scheme is infeasible, the proposed scheme can still work well.

Knowledge Graph



Sign up or login to leave a comment