Integer Precoding for FDD-Based MIMO Systems

How to achieve the closed-loop processing gain is an important issue for frequency division duplexing (FDD) based multi-input multi-output (MIMO) systems, since the channel state information (CSI) is not available by reciprocity as time division duplexing (TDD). Current technique mainly uses the codebook based precoding, where the receiver feeds back the precoding matrix index (PMI) of the codebooks to the transmitter for closed-loop processing. However, due to the suboptimality of linear processing and the quantization error of the precoding matrix / CSI, its performance gain over the open-loop processing is limited. In this paper, we propose a new precoding architecture to improve the situation. Based on lattice reduction (LR) method, we choose to feedback an integer matrix to the transmitter and perform nonlinear precoding. By utilizing the special properties of the integer matrix, the required feedback amount can be greatly reduced. Although it needs somewhat more bits than the conventional scheme to feedback the CSI, significant performance gain is achieved and full diversity can be guaranteed.