A Two-Stage Hybrid Beamforming Design for Full-Duplex mmWave Communications.

In this paper, we investigate the hybrid beamforming design for the full-duplex (FD) millimeter-wave (mmWave) multiple-input multiple-output (MIMO) system, aiming to maximize the overall spectral efficiency (SE) and cancel the self-interference (SI) simultaneously. The coupling transmit power constraints, unit-modulus constraints, and strong SI make the problem under study highly nonconvex. To solve the complicated problem, we first transform it into an equivalent yet more tractable weighted minimum mean square error minimization problem and then propose an effective two-stage hybrid beamforming design. In the first stage, we propose to extend the known matrix approximation method to the FD scenario to obtain the initial hybrid beamforming design. In the second stage, we propose to use the block coordinate descent method to update the digital beamformers and combiners while retaining the obtained analog solutions, which sufficiently exploits the easily adjustable property of digital solutions to cancel the residual SI and compensate for the SE loss. We further analyze the computational complexity of the proposed design. Numerical results demonstrate that our proposed two-stage hybrid beamforming design is superior to conventional related methods.