SHARQ: Scheduled HARQ for Time-and Loss-Rate-Sensitive Networks

Abstract:

5G networks provide high reliability and low delay thanks to Hybrid Automatic Repeat reQuest (HARQ). However, HARQ does not provide predictability for either reliability or delay and residual losses -mostly due to feedback losses and the non-adaptivity of the redundency versions- need to be compensated by the upper layers. We explore the possibility of extending HARQ to the transport layer with a cross-layer implementation that obtains the requirements from the application layer and the channel state information from the lower layers. We propose Scheduled HARQ (SHARQ), a capacity-approaching scheme that uses a graph-based algorithm to find the best repair schedule in polynomial time, thereby approaching the channel capacity much better than unscheduled codes.