A comparison of three adaptive transmission protocols for fountain-coded multicast

We investigate adaptive transmission for tactical packet radio networks in which a source uses fountain coding to multicast a file to a number of destinations over time-varying links. We examine two strategies from the literature for packet-by-packet adaptation of the modulation and channel code. The maximum data-recovery rate philosophy attempts to maximize the throughput for each transmitted packet. It can employ control information derived from practical receiver statistics. The weighted completion time approach minimizes a function of the expected file-delivery delays at the destinations and uses control information derived from perfect channel-state information and knowledge of the number of packets received by each destination. We devise a minimax completion time protocol that uses the same control information as the weighted completion time approach but applies the minimax criterion to the expected delays to select transmission parameters for each packet. Performance comparisons are given for a number of different multicast networks in which each link experiences correlated Rayleigh fading.