Analysis of Multi-User ARQ with Multiple Unicast Flows Under Non-iid Reception Probabilities

This paper presents an analysis of multi-user ARQ for multiple unicast flows with arbitrary reception probabilities. The core idea of multi-user ARQ is to exploit knowledge of previously overheard packets by unintended users, and perform wise selection and joint encoding of any retransmitted information. Intended users may decode the jointly encoded packet by exploiting the previously overheard packets. In this paper, a scheduling approach that balances the traffic shares and maximizes the throughput is proposed and analyzed. It is found that identical reception probabilities among the users yield the best performance and that the sensitivity in throughput efficiency, with respect to changes in reception probabilities, is reduced when the number of user increases.