Séminaire : Conception matérielle pour GRAND ; un décodeur universel pour les codes courts

Intervenant : Syed Mohsin ABBAS

Résumé : Guessing Random Additive Noise Decoding (GRAND) is a recently proposed universal Maximum Likelihood (ML) decoder for short-length and high-rate channel codes. GRAND is code-agnostic and noise-centric, which implies that, unlike other decoding techniques that use the structure of the underlying code to decode, GRAND attempts to guess the noise that corrupted the codeword during transmission through the communication channel. As a result, GRAND can decode any code, whether structured or unstructured. GRAND has hard-input and soft-input versions that are distinguished primarily by the order in which the channel-induced noise in the received signal is guessed. The focus of this talk is on designing high-throughput, area and energy efficient hardware architectures for GRAND and its variants. For applications with strict latency requirements and a tight energy consumption budget, designing area and energy efficient hardware is critical. The GRAND variants are analyzed in terms of decoding performance and computational complexity, and simplifications are proposed that result in efficient hardware implementations.