Parameter-Learned AMP for MIMO Signal Detection

Abstract

Approximate message passing (AMP) is applicable to large-scale MIMO signal detection and achieves a high detection performance with low computational complexity. However, the detection performance severely degrades when two conditions, i.e., the large-system limit and a channel matrix property that each element follows independent and identically distributed complex Gaussian distribution, are not satisfied. It has been found that the degradation is relaxed by introducing a constant multiplier to the observation rate which is the ratio of the numbers of received to transmitted signals. The optimal value of the multiplier depends on the numbers of transmit and receive antennas, signal-to-noise ratio, and other conditions. Recently, it has been proposed to optimize hyper parameter(s) by applying deep-unfolding concept to AMP. In this paper, two types of the learned AMP (LAMP) is proposed, and those detection performance is evaluated. The numerical evaluation results show that a modification to residual interference power is needed to optimize the hyper-parameters properly for the straightforwardly- implemented LAMP and that the higher performance is obtained by the simple structure LAMP in spite of omitting calculation of the residual interference power.