Decentralized Design of Fast Iterative Receivers for Massive MIMO with Spatial Non-Stationarities

Despite the extensive use of a centralized approach to design receivers for massive multiple-input multiple-output (M-MIMO) base stations, their actual implementation is a major challenge due to several bottlenecks imposed by the large number of antennas. One way to deal with this problem is by fully decentralizing the classical zero-forcing (ZF) receiver across multiple processing nodes based on the gradient descent (GD) method. In this paper, we first explicitly relate this decentralized receiver to a distributed version of the Kaczmarz algorithm and to the use of the successive interference cancellation (SIC) philosophy. In addition, we propose three methods to further accelerate the initial convergence of this decentralized receiver by exploring the connection with the Kaczmarz algorithm: 1) a new Bayesian distributed receiver, which iteratively estimates and eliminates noise; 2) a more practical method for choosing a relaxation parameter on an iteration basis; and 3) extension of it to tree-based architectures. The discussion also considers spatial non-stationarities that arise when the antenna arrays are extremely large (XL-MIMO). With our methods, we are able to improve the performance of the decentralized GD receiver for both spatially stationary and non-stationary channels, but mainly the non-stationary performance can still be improved. Future research directions are provided with the aim of further improving the effectiveness of receivers based on the so-called principle of successive residual cancellation (SRC).