Expectation Propagation Detection with Neumann-Series Approximation for Massive MIMO

Abstract

For massive multiple-input multiple-output (MIMO) systems, signal detection is always a key concern. Traditional detection methods such as minimum mean square error (MMSE) all suffer from a variety of problems in respect of complexity and performance when applied to large-scale MIMO systems. In this paper, expectation propagation (EP) algorithm is employed to guarantee high-accuracy and low-complexity of symbol detection in high-dimensional MIMO systems. Nevertheless, in the iterative updating process of EP algorithm, the inevitable matrix inversion operation is one of the key challenges to the realistic hardware implementation. Therefore, a new improved EP, which is termed as expectation propagation with Neumann-series approximation (EP-NSA), is firstly proposed to accommodate the complexity as well as the performance by executing an approximate matrix inversion with a small number of Neumann-series terms. Simulation results have shown that the proposed EP-NSA with two items achieves notable performance improvement compared to belief propagation (BP) detection when the system loading factor is relatively large. For antenna configurations with large loading factor, this approach achieves similar performance to MMSE while keeping lower computational complexity.