Sparse group vector coding for short-packet URLLC

Abstract Sparse vector coding (SVC) has become a promising short-packet transmission scheme, offering high reliability and low latency. Its transmission performance mainly depends on the construction of sparse vectors, namely sparse mapping. To further enhance performance, this paper proposes sparse group vector coding (SG-SVC), a novel transmission scheme featuring multi-constellation design with grouped allocation. Specifically, we first employ index modulation (IM) to allocate initial non-zero positions using a subset of index bits, and then determine extended non-zero positions through the remaining index bits. This methodology significantly mitigates the index table size problem inherent in conventional IM schemes. Afterwards, we group non-zero positions, assign distinct constellations to each group, and incorporate constellation swapping to enable additional bit transmission per group. Consequently, our scheme achieves both high transmission reliability and low latency. Simulation results demonstrate that the proposed SG-SVC outperforms existing SVC schemes in terms of transmission reliability and latency at high coding efficiencies.