A switched full duplex MIMO architecture with digital linear and nonlinear cancellation

Enabling full duplex (FD) in MIMO systems is challenging due to increased hardware complexity and increased training overhead required for canceling not only self-interference (SI) but also cross-link-interference (CLI) signals, considering both linear and nonlinear effects on each stream. In this paper, we propose switched FD MIMO (FD-SW-MIMO) architecture as a low-complexity, low-overhead solution, which enables stream-based nonlinear estimation to be performed independently from channel estimation, so that those nonlinear reference signals are fed to linear SI and CLI cancellation stages. For improved performance at high transmit power levels, the Random Fourier Features - Least Mean Squares (RFF-LMS) algorithm is employed on the residual SI and CLI signals per stream. Our experiments conducted on a software-defined radio based 2x2 FD MIMO test setup reveal that the proposed FD-SW-MIMO architecture can provide up to 12 dB enhancement over linear only digital cancellation. The proposed architecture requires only minor hardware modification(s), avoiding active analog cancellation circuitry and extra Tx/Rx chains. Requiring the same training overhead as linear only cancellation, FD-SW-MIMO architecture can quadruple the rate of HD SISO for low to moderate transmit power levels, and for high transmit power levels, the HD SISO rate is tripled due to slightly increased overhead.