Design and Decoding of Short Block Codes for URLLC Applications

Abstract

Ultra-reliable low-latency communication (URLLC) requires short-packet transmission under stringent reliability and latency constraints. In the short-block-length regime, coding schemes are evaluated against finite-block-length limits, such as the normal approximation (NA). Designing codes that approach these limits with low decoding complexity is particularly challenging at low code rates.

Primitive rateless (PR) codes are promising candidates for short-packet communications due to their strong distance properties, rate compatibility, and flexible block-length design. However, their practical use is limited by the high complexity of near-maximum-likelihood (ML) decoders, especially at low rates.

This thesis develops efficient decoding and coding architectures for PR codes. First, a double-bases belief propagation (BP) framework exploiting structured parity-check matrices derived from primitive polynomials is proposed. Building on this framework, a parallel hybrid decoder combining multiple-bases BP with low-order ordered statistics decoding (OSD) is developed, where parity-check matrices are optimized using a genetic algorithm. For very low-rate operation, a parallel concatenated coding scheme combining convolutional and PR codes is introduced. A parallel list Viterbi decoding strategy and an analytical approximation for the required list size are also proposed.

Simulation results show that the proposed schemes closely approach the NA benchmark while significantly reducing complexity compared with high-order OSD. Moreover, they outperform several state-of-the-art short block-length coding schemes, particularly in the very low-rate regime, demonstrating the potential of PR codes for URLLC applications.