An Improved Weight Design for Unwanted Packets in Multicast Instantly Decodable Network Coding

Abstract

Traditional studies on Instantly Decodable Network Coding (IDNC) focus on broadcast scenarios in which all users are interested in receiving the same packets. The problem is usually formulated as a maximum weight clique search in the IDNC graph for well-designed weights that reflect the metric under investigation, i.e., completion time or decoding delay. These schemes are allegedly extended to multicast sessions by introducing a two-layered IDNC graph wherein the first layer contains Wanted Packets (WPs), and the second one represents UnWanted Packets (UWPs). This letter studies the effect of UWPs on the performance of IDNC schemes for multicast sessions. In particular, the letter derives the condition under which the two-layer structure is optimal and concludes that the approach is the most appropriate in most, but not all, cases. Furthermore, the letter designs weights for UWPs that do not depend on the metric but rather represent their probability of offering future decoding opportunities for WPs. Numerical results reveal that the proposed solution outperforms existing methods for an identical computation complexity.