An Efficient Retransmission Based on Network Coding with Unicast Flows

Recently, network coding technique has emerged as a promising approach that supports reliable transmission over wireless loss channels. In existing protocols where users have no interest in considering the encoded packets they had in coding or decoding operations, this rule is expensive and inef-ficient. This paper studies the impact of encoded packets in the reliable unicast network coding via some theoretical analysis. Using our approach, receivers do not only store the encoded packets they overheard, but also report these information to their neighbors, such that users enable to take account of encoded packets in their coding decisions as well as decoding operations. Moreover, we propose a redistribution algorithm to maximize the coding opportunities, which achieves better retransmission efficiency. Finally, theoretical analysis and simulation results for a wheel network illustrate the improve-ment in retransmissions efficiency due to the encoded packets.

💡 Research Summary

The paper tackles the classic problem of reliable unicast transmission over lossy wireless links by exploiting network coding in a more informed manner than prior work. Traditional network‑coding schemes such as COPE assume that receivers simply discard any encoded packet they overhear unless they can immediately decode it; consequently, the encoded packet is treated as a “by‑product” rather than a useful resource. This assumption limits coding opportunities and forces many unnecessary retransmissions.

The authors propose a two‑fold innovation. First, each receiver stores every overheard encoded packet in a local buffer and periodically reports a concise metadata list (packet identifiers and the set of native packets it already possesses) to its neighbors, especially the central transmitter in a wheel topology. This “encoded‑packet reporting” turns the encoded packet into a shared state that the transmitter can query when making coding decisions. Second, they introduce a redistribution (or re‑allocation) algorithm that uses the collected state information to maximize simultaneous decoding opportunities. The algorithm models the network as a weighted graph where vertices represent receivers and edges indicate the existence of a common encoded packet that could serve both endpoints. Edge weights reflect how many receivers would benefit from a particular XOR combination. By solving a maximum‑weight matching (or flow) problem on this graph, the transmitter selects a set of packet pairs (or larger groups) that can be XOR‑encoded together, guaranteeing that each selected receiver can later decode using either its own native packets or other encoded packets it has already stored.

Theoretical analysis focuses on a wheel network consisting of a single hub and N leaf nodes, each link experiencing an independent loss probability p. Without reporting, the expected number of retransmissions per original packet is 1/(1‑p). With the reporting mechanism, the authors derive an expected retransmission count of

 E