In this paper, an index-coded Automatic Repeat Request (ARQ) is studied in the perspectives of transmission efficiency and memory overhead. Motivated by reducing significant computational complexity from huge matrix inverse computation of random linear network coding, a near-to-optimal broadcasting scheme, called index-coded Automatic Repeat Request (ARQ) is proposed. The main idea is to consider the principal packet error pattern across all receivers. With the help of coded side information formed by successfully decoded packets associated with the dominant packet error pattern, it is shown that two contradictory performance metrics such as transmission efficiency and transmit (receive) cache memory size for index coding (decoding) can be enhanced with a reasonable trade-off. Specifically, the transmission efficiency of the proposed scheme is proved to be asymptotically optimal, and memory overhead is shown to be asymptotically close to the conventional ARQ scheme. Numerical results also validate the proposed scheme in the sense of memory overhead and transmission efficiency in comparison with the conventional ARQ scheme and the optimal scheme using random linear network coding.
Keywords: Automatic Repeat Request (ARQ); index coding; random linear network coding.