Multicast Multiuser ARQ
Peter Larsson · 2008
In this paper, multiuser ARQ is extended to multicasting. The core idea is that the sender, based on feedback from users regarding successfully received transmissions, adapts code weights for data packet linear combinations that are then sent. Each user exploits its previously received information in decoding the linearly combined packets. Specifically, a throughput optimal, low en-/decoding complexity enabling, low overhead and on-line multicast coding and scheduling algorithm is devised based on a per user rank increase criterion. For throughput optimality, a minimum field size criterion is derived. Relative previous work, which adaptively identifies sets of users suited to receive linearly combined packets and uses GF(2) and XOR coding, the proposed method adaptively select weights from a sufficient large finite field for optimality instead. Throughput is analyzed and simulated, and en-/decoding complexity, signaling overhead, and latency etc. are studied through realistic simulations. Overall, it is found that the throughput is significantly higher than multicast selective repeat ARQ, and that the optimal throughput for an erasure channel is attained.