Complexity Reduction in AMP Iterative Detection: A New Approach With Error Function-Aided Mechanism and Convergence-Based Termination

TY - JOUR

T1 - Complexity Reduction in AMP Iterative Detection

T2 - A New Approach With Error Function-Aided Mechanism and Convergence-Based Termination

AU - Li, Huan

AU - Huang, Jingxuan

AU - Fei, Zesong

AU - Guo, Jing

AU - Yuan, Weijie

N1 - Publisher Copyright: © 2002-2012 IEEE.

PY - 2026

Y1 - 2026

N2 - Approximate message passing (AMP) iterative detection is recognized as a reliable and practical approach for multiple-input multiple-output (MIMO) systems. However, existing AMP detection algorithms face a critical challenge: high computational complexity due to redundant iterations, making them impractical for the coming 6G networks with increased data throughput demands. This paper addresses this challenge by investigating the mutual information (MI) update flow in AMP iterative MIMO detection and introducing a precise MI computation mechanism based on the error function, referred to as the EFA mechanism. Leveraging the EFA mechanism, we propose a convergence-based termination (CT) scheme to accurately track the convergent iteration number and eliminate redundant iterations in AMP iterative detection. Numerical results demonstrate that the MI flow calculated using the EFA mechanism is consistent with the convergence behavior of AMP iterative MIMO detection across different iterations and signal-to-noise ratios (SNRs). Specifically, the EFA mechanism can precisely identify the convergent iteration number and corresponding SNR. Additionally, the CT scheme achieves up to an 80% reduction in complexity compared to the original AMP detection, while maintaining the expected BER performance.

AB - Approximate message passing (AMP) iterative detection is recognized as a reliable and practical approach for multiple-input multiple-output (MIMO) systems. However, existing AMP detection algorithms face a critical challenge: high computational complexity due to redundant iterations, making them impractical for the coming 6G networks with increased data throughput demands. This paper addresses this challenge by investigating the mutual information (MI) update flow in AMP iterative MIMO detection and introducing a precise MI computation mechanism based on the error function, referred to as the EFA mechanism. Leveraging the EFA mechanism, we propose a convergence-based termination (CT) scheme to accurately track the convergent iteration number and eliminate redundant iterations in AMP iterative detection. Numerical results demonstrate that the MI flow calculated using the EFA mechanism is consistent with the convergence behavior of AMP iterative MIMO detection across different iterations and signal-to-noise ratios (SNRs). Specifically, the EFA mechanism can precisely identify the convergent iteration number and corresponding SNR. Additionally, the CT scheme achieves up to an 80% reduction in complexity compared to the original AMP detection, while maintaining the expected BER performance.

KW - Gaussian approximation

KW - MIMO detection

KW - Mutual information

KW - convergence

KW - early termination

KW - factor graph

UR - https://www.scopus.com/pages/publications/105010700319

U2 - 10.1109/TWC.2025.3585498

DO - 10.1109/TWC.2025.3585498

M3 - Article

AN - SCOPUS:105010700319

SN - 1536-1276

VL - 25

SP - 602

EP - 617

JO - IEEE Transactions on Wireless Communications

JF - IEEE Transactions on Wireless Communications

ER -