A Dynamic-Confined Iterative GRAND Algorithm With Anchor Decoding for Product Codes

Abstract

In this letter, we propose two novel low-complexity iterative guessing random additive noise decoding (IGRAND) schemes for product codes with enhanced anchor decoding (EAD). Both schemes are motivated by the observation that errors typically occur at the intersections of rows and columns that declare decoding failure by bounded distance decoding (BDD). For the first decoding scheme, termed confined IGRAND with EAD, the generation of test error patterns (TEPs) for the component GRAND is strictly confined to the intersections of invalid codewords which are detected before decoding. To further reduce the decoding complexity, we propose the second decoding scheme where the constraint on the generation of TEPs is dynamically updated according to the hard reliability scores (HRSs) in EAD. Specifically, the codewords are classified into reliable and unreliable subsets by a threshold and the generation of TEPs is strictly confined to the intersections of unreliable codewords during the decoding process. Such a decoding algorithm is referred to as dynamic-confined IGRAND with EAD. Extensive simulation results show that both decoding schemes achieve significant complexity reduction with only negligible performance loss. These results confirm the potential of the proposed decoding schemes in high-throughput applications.