Optimization of Uplink CSI Training for Full-Duplex Multiuser MIMO Systems

In this letter, we investigate the optimization of uplink (UL) channel state information (CSI) training in the full-duplex (FD) based multiuser (MU) multiple-input multiple-output (MIMO) systems. Here, an FD-assisted BS performs simultaneous UL CSI training and downlink data transmission that produces self interference (SI). Following the UL training stage, the BS uses the conventional half-duplex (HD) MU MIMO transmission with zero-forcing precoding. To find an optimal UL CSI training length that achieves a balance between the training overhead and CSI quality, we formulate an optimization problem that maximizes the sum spectral efficiency of the network. To ease the analysis, we derive a lower bound on the user rate in the FD phase, which is then used together with the HD user rate to obtain the sub-optimal solution. We also provide a closed-form expression to approximate the UL training length. Numerical results show that the performance of the proposed UL training outperforms the fixed length training and it closely matches the performance with an exhaustive search.