Linear Transceiver Design for Bidirectional Full-Duplex MIMO OFDM Systems

In this paper we address the linear precoding and decoding design problem for a bidirectional orthogonal-frequency-division-multiplexing (OFDM) communication system, between two multiple-input-multiple-output (MIMO) full-duplex (FD) nodes. The effects of hardware distortion, leading to the residual self-interference and inter-carrier leakage, as well as the channel state information (CSI) error are taken into account. In the first step, the operation of a FD MIMO OFDM transceiver is modeled where the explicit impact of hardware inaccuracies on the inter-carrier leakage is observed. An alternating quadratic convex program (AltQCP) is then provided to obtain a minimum-mean-squared-error (MMSE) design for the defined system. Moreover, taking into account the impacts of CSI inaccuracy, an alternating semi-definite-program (AltSDP) is proposed to obtain a worst-case MMSE design under a norm-bounded CSI error. The provided designs are also extended to maximize the system sum rate, applying the weighted-MMSE (WMMSE) method. The proposed AltQCP and AltSDP algorithms are based on alternating update of the optimization variables, with a guaranteed convergence, where in each step the sub-problem is solved to optimality. In order to provide further insights, the computational complexity of the AltSDP algorithm is analytically obtained in relation to the problem dimensions. Moreover, a methodology to obtain the least favorable CSI error matrices is obtained, by transforming the resulting non-convex quadratic problem into a convex problem. Finally, the proposed methods are numerically evaluated in terms of the achievable system performance, computational complexity, as well as the comparison to other approaches in the literature. A significant gain is observed via the application of the proposed methods as the hardware inaccuracy, and consequently inter-carrier leakage, increases.