Improved Soft-Input Soft-Output Turbo-GRAND Data Detection for MIMO Wireless Systems

Guessing random additive noise decoding (GRAND) is a code-agnostic maximum likelihood (ML) decoding approach that attempts to recover the received codewords by identifying noise patterns. Compared to decoding alone, iterative detection and decoding can significantly improve the frame error rate (FER) performance of wireless communication systems by exchanging soft information between the data detector and the channel decoder. However, the existing GRAND-centered iterative detection and decoding framework, called turbo-GRAND, suffers unsatisfactory FER performance due to the less reliable log-likelihood ratio (LLR) computation at the detection stage and the worse ability to determine valid codewords at the decoding stage. In this letter, we propose an improved soft-input soft-output (SISO) turbo-GRAND data detection for multiple-input multiple-output (MIMO) wireless systems, in which we introduce the low-complexity bit-flip and codeword generation strategies to further boost the FER performance. Evaluated in different scenarios for MIMO wireless systems that are coded by a (128, 105) polar code, simulation results demonstrate that our improved turbo-GRAND algorithm outperforms the original turbo-GRAND by Sarieddeen et al., GLOBECOM 2022 by about 3.4 dB to 4.5 dB at the target FER of 10-3.