Studies on the performance of some ARQ schemes

Abstract

This thesis consists of a summary part and seven published articles. All the articles are about performance analysis of ARQ schemes.

Two of the publications study the performance of an ARQ scheme with packet combining, called the EARQ (extended ARQ) scheme. In the packet combining algorithm, the bitwise modulo-2 sum of two erroneous copies of a packet is computed to locate the errors. The packet combining algorithm involves a straightforward search procedure, the computational complexity of which easily becomes prohibitive. As a solution to this, a modified scheme is proposed, where the search procedure is attempted only when there are at most Nmax 1s at the output of the modulo-2 adder. In one article, time diversity was utilized, whereas space diversity reception was considered in the other work.

The remaining five publications study the throughput performance of adaptive selective-repeat and go-back-N ARQ schemes, where the switching between the transmission modes is done based on the simple algorithm proposed by Y.-D. Yao in 1995. In this method, α contiguous NACKs or β contiguous ACKs indicate changes from 'good' to 'bad' or from 'bad' to 'good' channel conditions, respectively. The numbers α and β are the two design parameters of the adaptive scheme. The time-varying forward channel is modelled by two-state Markov chains, known as Gilbert-Elliott channel models. The states are characterized by bit error rates, packet error rates or fading parameters. The performance of the adaptive ARQ scheme is measured by its average throughput over all states of the system model, which is a Markov chain. A useful upper bound for the achievable average throughput is provided by the performance of an (assumed) ideal adaptive scheme which is always in the 'correct' transmission mode. The optimization of α and β is done based on minimizing the mean-square distance between the actual and the ideal performance curves. Methods of optimizing the packet size(s) used in the adaptive selective-repeat scheme are also proposed.