Graph Neural Network Aided Expectation Propagation Detector for MU-MIMO Systems

TL;DR

This paper introduces GEPNet, a novel detector combining expectation propagation and graph neural networks to improve detection performance in high-interference MU-MIMO systems, surpassing existing methods.

Contribution

The paper presents a new GEPNet detector that overcomes limitations of traditional EP by integrating graph neural networks for better interference handling in MU-MIMO.

Findings

GEPNet outperforms state-of-the-art detectors in strong MUI scenarios.

The proposed method effectively addresses the Gaussian approximation limitation in EP.

Simulation results demonstrate significant performance gains.

Abstract

Multiuser massive multiple-input multiple-output (MU-MIMO) systems can be used to meet high throughput requirements of 5G and beyond networks. In an uplink MUMIMO system, a base station is serving a large number of users, leading to a strong multi-user interference (MUI). Designing a high performance detector in the presence of a strong MUI is a challenging problem. This work proposes a novel detector based on the concepts of expectation propagation (EP) and graph neural network, referred to as the GEPNet detector, addressing the limitation of the independent Gaussian approximation in EP. The simulation results show that the proposed GEPNet detector significantly outperforms the state-of-the-art MU-MIMO detectors in strong MUI scenarios with equal number of transmit and receive antennas.