Performance of CA-Polar Codes Serial Concatenated with RLL Code Using SCL-SGRAND on VLC

The Guessing Random Additive Noise Decoding (GRAND) algorithm is a well-established hard-decision technique that achieves maximum likelihood (ML) decoding performance. A recent variant, Soft-GRAND (SGRAND), can utilize soft information as input and is applicable to various short, high-rate block codes. We propose a scheme called successive-cancellation list (SCL)-SGRAND, where SCL decoding is employed to generate both soft information (reliability estimates) and initial hard bit decisions. These decisions are then improved using SGRAND when they fail to satisfy the cyclic redundancy check (CRC) condition. This combined approach allows SCL-SGRAND to outperform both the original SCL and SGRAND decoders. To evaluate its performance, we compare SCL-SGRAND's performance on CRC-aided Polar (CA-Polar) codes, serially concatenated with Run-Length Limited (RLL) codes (a common configuration for Visible Light Communication (VLC)), against SCL decoding and SGRAND under identical channel conditions (e.g., AWGN channel). The simulation results demonstrate that the proposed SCL-SGRAND scheme outperforms both SCL and SGRAND, achieving gains of between 0.2 and 1 dB for bit error rate (BER) and between 0.3 and 0.6 dB for block error rate (BLER).