A Low-Complexity ADMM-based Massive MIMO Detectors via Deep Neural Networks

TL;DR

This paper introduces a deep neural network approach to improve ADMM-based massive MIMO detectors by learning optimal penalty parameters, resulting in enhanced performance and reduced complexity for high-order modulation signals.

Contribution

It proposes a novel DNN-based detector that unfolds ADMM and learns penalty parameters, overcoming traditional limitations and enabling efficient detection in large-scale MIMO systems.

Findings

The DNN-based detector outperforms traditional ADMM detectors in accuracy.

The proposed method effectively handles higher-order modulation signals.

A computationally cheaper detector with comparable performance is also introduced.

Abstract

An alternate direction method of multipliers (ADMM)-based detectors can achieve good performance in both small and large-scale multiple-input multiple-output (MIMO) systems. However, due to the difficulty of choosing the optimal penalty parameters, their performance is limited. This paper presents a deep neural network (DNN)-based massive MIMO detection method which can overcome the above limitation. It exploits the unfolding technique and learns to estimate the penalty parameters. Additionally, a computationally cheaper detector is also proposed. The proposed methods can handle the higher-order modulation signals. Numerical results are presented to demonstrate the performances of the proposed methods compared with the existing works.