GRAND Massive Parallel Decoding Framework for Low Latency in Beyond 5G

TL;DR

This paper introduces a massively parallel decoding framework called GRAND for low-latency applications in beyond 5G, utilizing novel likelihood functions and matrix operations to efficiently decode a wide range of codeword lengths.

Contribution

The paper presents a new parallel decoding framework that significantly reduces decoding complexity and latency for 5G NR codes and modulation schemes.

Findings

Reduces symbol error pattern space from O(5^{N/log_2 M}) to O(4^{N/log_2 M})

Performs matrix-vector multiplication in O(log_2 N) steps

Effective for codewords from 32 to 1024 bits in 5G NR

Abstract

We propose a massive parallel decoding GRAND framework. The framework introduces two novelties: 1. A likelihood function for $M$-QAM demodulated signals that effectively reduces the symbol error pattern space from $\mathcal{O}(5^{N/\log_2 M})$ down to $\mathcal{O}(4^{N/\log_2 M})$; and 2. A massively parallel matrix-vector multiplication for matrices of size $K\times N$ ($K \leq N$) that performs the multiplication in just $\mathcal{O}(\log_2 N)$ steps. We then apply the proposed GRAND approach to codes and operational modulation techniques used in the current 5G NR standard. Our framework is applicable not just to short codewords but to the full range of codewords from 32 bits up to 1024 bits used in the control channels of 5G NR. We also present simulation results with parity-check matrices of Polar codes with rate $R=1/2$ obtained from the 5G NR universal reliability sequence.