Conceitos Básicos
A highly parallelizable decoding framework based on the Guessing Random Additive Noise Decoding (GRAND) approach that can efficiently process higher-order modulation techniques used in 5G NR control channels.
Resumo
The paper proposes a massive parallel decoding framework using a GRAND-like approach, focusing on extensive parallelization to achieve low latency in beyond 5G networks. The key highlights are:
The framework introduces a likelihood function for M-QAM demodulated signals that effectively reduces the symbol error pattern space from O(5N/log2 M) to O(4N/log2 M).
It describes a novel massively parallel matrix-vector multiplication algorithm that performs the multiplication in just O(log2 N) steps. This is applied to the parity-check matrices of Polar codes used in 5G NR.
The proposed approach is evaluated for all block lengths (N = 32, 64, 128, 256, 512, 1024 bits) specified for use in the 5G NR control channels and various M-QAM modulation schemes (M = 4, 16, 64, 256, 1024, 4096).
Simulation results show that the proposed GRAND-like approach provides good block error rate (BLER) performance across the different M-QAM schemes and block lengths, while achieving the goal of low latency decoding.
Unlike other GRAND approaches that aim to achieve SNR gains, this work focuses exclusively on maximizing parallelization to reduce latency, which is a key requirement for beyond 5G networks.
Estatísticas
The size of the symbol error pattern space in the proposed approach is O(4N/log2 M).
The proposed parallel matrix-vector multiplication algorithm takes 1 + max(i∈1,...,K⌈log2(WH(i))⌉) parallel steps, where WH is the total Hamming weight of the parity-check matrix H.
Citações
"We introduce a likelihood function for M-QAM demodulated signals that effectively reduces the symbol error pattern space from O(5N/log2 M) down to O(4N/log2 M)."
"We describe a novel massively parallel matrix-vector multiplication that performs the multiplication in just O(log2 N) steps."