Reliability-based hybrid ARQ using convolutional codes

Abstract

Reliability-based hybrid ARQ (RB-HARQ) is a recently introduced approach to incremental-redundancy ARQ. In RB-HARQ, the bits that are to be retransmitted are adaptively selected at the receiver based on the estimated bit reliabilities. This technique has the potential to improve performance and minimize retransmission size by targeting those bits that are likely to be in error. However, previous versions of the RB-HARQ algorithm have resulted in large request (NACK) packets that made the technique appropriate only for systems that can tolerate large request packets on the feedback link. In this paper, we show that RB-HARQ is effective with convolutional codes, and we exploit the time-correlation properties of these codes to significantly reduce the size of the retransmission requests.