Soft-Decision Randomized Likelihood Decoding

Ultra-reliable low-latency communications (URLLC) scenario for 6G drives the interest in studying and development of existing and new universal decoding algorithms capable of providing near-optimum error-correcting performance for short block length codes with ultimate goal of providing low latency and moderate complexity. In this paper we consider a new universal decoding technique based on randomized likelihood (RL) rule that may be implemented for practical use in the future. U sing this rule we get the message estimate as a single sample of the message vector from a posterior distribution of messages given the received signal. A soft-decision implementation of RL decoding algorithm is suggested. After suggesting further complexity optimization, the new algorithm is compared to other well-known universal decoders, such as ordered-statistics decoding (OSD) and guessing random additive noise decoding (GRAND). The simulation results for short codes with different rates are provided for binary-input additive white Gaussian noise (BAWGN) communication channel.